You are on page 1of 96

ENGO 421: COORDINATE SYSTEMS

Alexander Braun
Department of Geomatics Engineering
Schulich School of Engineering
University of Calgary

2008
Contents
1 Introduction 4
1.1 What is geodesy and why coordinate systems? . . . . . . . . . . . . . . . . . 5
1.2 Geodetic measurements and errors . . . . . . . . . . . . . . . . . . . . . . . 9

2 System of Natural Coordinates 13


2.1 Newton’s Law of Gravitation - Gravitational Acceleration and Potential . . . 13
2.2 Gravity Potential and Centrifugal Potential . . . . . . . . . . . . . . . . . . . 19
2.3 Level Surfaces and Plumb Lines . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Natural Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3 Celestial Coordinate Systems 30


3.1 The Celestial Sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2 The Horizon System - H . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 The Right Ascension System - RA . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4 The Hour Angle System - HA . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.5 The Ecliptic System - E . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.6 Reflection and Rotation Matrices . . . . . . . . . . . . . . . . . . . . . . . . 39
3.7 Transformations Between Celestial Systems . . . . . . . . . . . . . . . . . . 40
Transformation between HA and RA . . . . . . . . . . . . . . . . . . . . . . 40
Transformation between RA and E . . . . . . . . . . . . . . . . . . . . . . . 42
Transformation between H and HA . . . . . . . . . . . . . . . . . . . . . . . 42
Other transformations between H, RA, HA, and E . . . . . . . . . . . . . . . 43
3.8 Astronomical Triangle and Spherical Trigonometry . . . . . . . . . . . . . . 43
Spherical triangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Astronomical triangle connecting H and HA . . . . . . . . . . . . . . . . . . 46
Time systems used in astronomic azimuth determination . . . . . . . . . . . 47
Sidereal time and hour angle . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Determination of the astronomic azimuth . . . . . . . . . . . . . . . . . . . 48
3.9 Hour angle method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
The error budget for the hour angle method using Polaris: . . . . . . . . . . 50
Observation procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Computation procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.10 Astronomic azimuth determination by the altitude method . . . . . . . . . . 52
Observation procedure for azimuth determination by altitude observations
of the Sun . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

C Alexander Braun (2006-2010)


2 ENGO421: COORDINATE SYSTEMS
Computation procedure for azimuth determination by altitude observations
of the Sun . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

4 Time Systems 55
4.1 Sidereal Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2 Solar or Universal Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.3 Conversion between Sidereal and Solar Time . . . . . . . . . . . . . . . . . 61
4.4 Time Zones and Calendar Time . . . . . . . . . . . . . . . . . . . . . . . . . 62
Time Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Calendar Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.5 Atomic Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Time Transmitters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

5 Terrestrial Coordinate Systems 66


5.1 The Best Fitting Ellipsoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.2 Basic Ellipsoidal Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.3 Coordinates on the Ellipsoid - Geodetic Coordinate Systems . . . . . . . . . 72
Geodetic Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
The Reduced Latitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
The Geocentric Latitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Relations Between Φ, φ, β, γ . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.4 Transformation Between Cartesian and Curvilinear Geodetic Coordinates . . 80

6 Coordinate Transformations 84
6.1 Transformation Between Systems with Different Origins and Orientations . . 85
6.2 Transformations Between Local and Global Systems . . . . . . . . . . . . . . 87
6.3 The Datum Problem Today . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.4 Summary of Coordinate Systems and Transformations . . . . . . . . . . . . 93

7 Basics of Map Projections 93


7.1 Direct and inverse geodetic problem . . . . . . . . . . . . . . . . . . . . . . 94
7.2 Coordinate transformations in mapping . . . . . . . . . . . . . . . . . . . . 95

C Alexander Braun (2006-2010)


3 ENGO421: COORDINATE SYSTEMS
Preface
These lecture notes have been developed during the course ENGO 421 in Fall 2006. In-
dividual chapters will be subsequently distributed to the students through the Blackboard
course management system at http://blackboard.ucalgary.ca.
These lecture notes are based on Dr. K.-P. Schwarz’s lecture notes on Fundamentals of
Geodesy, last published in 1999. The contents are, however, modified in order to incorpo-
rate new information and knowledge as well as having a rearranged order of topics to ac-
commodate for the required Astronomical Observations Lab, which now takes place in the
Fall term and not in the Spring term anymore. Due to the different seasonal weather con-
ditions, the chapter on celestial coordinate systems has been placed before the terrestrial
coordinate systems to give students the required background to carry out the astronomical
observations early in the term. Additional material was provided by Dr. Nico Sneeuw,
Dr. Michael G. Sideris, and Dr. Rossen S. Grebenitcharsky. Their contributions are highly
appreciated.
For the Fall 2008 term, the order of chapters in class was modified for scheduling reasons.
The lab assignment on astronomic azimuth determination requires to cover chapters 3 and
4 before chapter 2 in class, the original order is, however, kept in the notes.

1 Introduction
The Earth is a dynamic planet which changes its form, composition and location constantly.
In order to quantify the shape, the deformation and the distribution of masses (rock, wa-
ter, snow and ice) on Earth, which is mostly described by the gravity field of the Earth,
the discipline of geodesy is employed. One of the key aspects of geodesy is to establish
coordinate systems and reference systems which can be used to consistently describe the
shape, the deformation and the gravity field. This course is designed to make student
familiar with the fundamentals of coordinate systems which are frequently used in geo-
matics engineering and geodesy. The student will develop an understanding of terrestrial
and celestial coordinate systems, the transformation between systems, and the geodetic
principles of map projections.
In order to understand these topics, the fundamentals of geodesy and particularly the
gravity field are required. These include natural coordinates (Chapter 2), gravitational
and centrifugal acceleration and force, gravity and gravity potential, geoid, plumb lines
and equipotential surfaces. As the Earth is of complex shape and mass distribution, a
mathematical approximation of the above parameters is thought after. These include nor-
mal gravity, ellipsoid and geodetic coordinates. A description of the Earth’ motion in space
requires a non-terrestrial coordinate system, hence, celestial coordinate systems (Chapter
3) are introduced including the celestial sphere, Horizon system, right ascension system,
hour angle system, ecliptic system, astronomical positioning and time systems (Chapter
4).
Terrestrial coordinate systems (Chapter 5) must be discussed including astronomic and
geodetic coordinates, the transformation of Cartesian and curvilinear coordinates, merid-
ians and parallels, and the geodesic. Once the terrestrial and celestial coordinate systems
have been developed, the transformation tools are discussed which allow to transform
coordinates (Chapter 6) from one system to any other system. This concerns also the
datum problem and the sensors providing information for the establishment of reference
systems. The final chapter will elaborate on map projections (Chapter 7) and particu-

C Alexander Braun (2006-2010)


4 ENGO421: COORDINATE SYSTEMS
larly the effects involved in the projection of coordinates from the 3-dimensional to the
2-dimensional space. This concerns Tissot’s indicatrix, map distortions, and differential
geometry on the sphere and ellipsoid.
A sketchy tour through the course

The continuation of this course will take place in ENGO 423: Geodesy, which will be taught
in the Winter term. This course will focus on physical geodesy and more details on Earth
rotation and tides, dynamic coordinate systems and a more sophisticated treatment of the
gravity field.

1.1 What is geodesy and why coordinate systems?


Geodetic or geomatics measurements take place in a natural environment, which shows
spatio-temporal variations. The natural effects influence the measurements and need to be
corrected before an analysis can take place. After this treatment of the measurements, the
use of coordinate systems is required to represent the results of such measurements and
to make them available to users. To capture the essential characteristics of a discipline,
definitions are often a good starting point. The list below gives an extensive though not
exhaustive collection of definitions of geodesy. The first two, venerated by age, mark in a
way two extreme positions. The first one is due to Bruns (1878), one of the most creative
geodesists of the 19th century. He states that “the task of geodesy is the determination of
the potential function W(x, y, z)”, i.e. of the gravity potential of the Earth. The second
one is due to Helmert (1880), one of the towering figures in geodesy around the turn of
the 19th century. He states that “geodesy is the science of measuring and mapping the
Earth’s surface”. At a first look, these two definitions seem unrelated. At a deeper level
though, they represent two sides of the same coin. They indicate that positioning and
gravity field determination are really not separate tasks, but need to be treated together.
However, for practical purposes, we often look at them individually. A definition which
incorporates both points of view was published in 1973 by the National Research Council
of Canada “Geodesy is the discipline that deals with the measurement and representation
of the Earth, including its gravity field, in a three dimensional time varying space”. We
will use this definition to outline some of the fundamental questions that are at the core
of geodesy.
Coordinates provide information about the location or position of objects, coordinates
represent the language in navigation, positioning, and mapping. Without coordinates,
measurements do not provide sufficient information for others to fully comprehend and
reproduce a position or location. Coordinates alone are not sufficient as well, as they
require a reference which is used to relate the coordinates to. These reference systems
represent the dictionary which everybody can use to look-up what a coordinate means
with respect to a certain reference system. Finally, there are different languages and dif-
ferent dictionaries, but translators allow us to communicate between different systems or
to exchange coordinates between different systems. The translator in this case are the
coordinate transformation tools to be developed in this course. In principle, we learn the
language of geodesy in this course which enables us to communicate measurements across
countries and disciplines. The key points of establishing geodetic information or models is
thus:

1. information/measurements

C Alexander Braun (2006-2010)


5 ENGO421: COORDINATE SYSTEMS
2. a reference system in space and time

C Alexander Braun (2006-2010)


6 ENGO421: COORDINATE SYSTEMS
C Alexander Braun (2006-2010)
7 ENGO421: COORDINATE SYSTEMS
3. measured corrections to correct information, reductions, systematic errors, bias
4. models of corrections, if measurements are not available
5. projection parameters to create maps
6. interpretation of measurements/maps and application

DEFINITIONS OF GEODESY

1. The task of geodesy is the determination of the potential function W(x,y,z). Bruns,
1878
2. Geodesy is the science of measuring and mapping the Earth’s surface. Helmert, 1880
3. Geodesy is a branch of science which investigates methods to accurately measure
elements of the Earth’s surface and to determine from them geographic positions of
points on this surface and which studies the figure of the Earth from a theoretical
point of view and by evaluating results of measurements. Zakatov, 1957
4. Geodesy is both theoretical and practical. Its theoretical function is to determine
the size and shape of the Earth and, in conjunction with other Earth sciences, to
study the structure of the Earth crust and of the immediately underlying layers. Its
practical function is to perform the measurements and computations that will give
the coordinates of selected control points on the Earth’s surface, i.e., to fix their
positions on the Earth’s surface. Heiskanen and Vening Meinesz, 1958
5. Au sens étymologique du mot, la géodésie est la science qui a pour objet la mesure
des dimensions de la Terre. Déterminer, d’une part, la forme et les dimensions
précises de la plante; réaliser, d’autre part, principalement au moyen de triangula-
tions, la mensuration des territoires terrestres pour permettre d’endresser des cartes
exactes et fournir des données géométrique précises pour les diverses enterprises de
l’ingénieur, sont en effet les buts principaux, scientifiques et practiques de l’activité
des géodésiens. Dupuy and Dufour, 1969
6. Geodesy is a discipline that deals with measurement and representation of the Earth,
including its gravity field, in a three-dimensional time varying space. NRC, 1973
(Vanicek and Krakiwsky, 1982)
7. Geodesy is considered as a discipline which deals mainly with the mapping of the
Earth and the monitoring of variations at its surface. From the very beginning
those tasks were connected with the gravity vector ~g (absolute value and direction).
Groten, 1979
8. The problem of geodesy is to determine the figure and the external gravity field of
the Earth and of other heavenly bodies as functions of time; as well as to determine
the mean Earth ellipsoid from parameters observed on and exterior to the Earth’s
surface. Torge, 1980
9. Theoretical Geodesy is that part of geodesy which has as its task the solution of
scientific problems of geodesy - the determination of the figure of the Earth and its
external gravity field, as well as their temporal variations - by means of geodetic
measurements. Pellinen/Deumlich, 1981

C Alexander Braun (2006-2010)


8 ENGO421: COORDINATE SYSTEMS
10. Geodesy is the scientific discipline that deals with the measurement and representa-
tion of the Earth, its gravitational field and geodynamic phenomena (polar motion,
earth tides, and crustal motion) in three-dimensional, time-varying space (Wikipedia,
http://en.wikipedia.org/wiki/Geodesy (last accessed, Sep 2006).
Today, the field of geodesy becomes more and more interdisciplinary and touches disci-
plines such as geomatics, remote sensing, geophysics, oceanography, hydrology, glaciol-
ogy, environmental studies, atmospheric and space science. It represents the foundations
of geomatics engineering and provides most of its disciplines with the tools to handle and
communicate measurements about the Earth’ surface, its interior, its dynamics, and the
gravity field.

1.2 Geodetic measurements and errors


Measurements take place in the physical world and include errors and uncertainties.
Geodesy deals with measurements and many of its problems are related to their avail-
ability, accuracy, resolution, and distribution. Measurements can never be made without
errors. It is therefore important to distinguish between the measurement and the quantity
to be measured. The first is often called observation, the second observable. Thus, an
observation or measurement is the observable plus errors. Take as an example the mea-
surement of a distance between two points A and B. The observable is in this case the
distance itself while its measurement contains errors of systematic and stochastic nature.
For instance, if a tape is used, systematic errors arise from the scale factor and tape length
variations due to temperature or tension. These systematic errors can be modeled or cali-
brated. Stochastic errors are due to the inaccuracy of reading the tape and to insufficient
control over the environmental conditions. Although these errors cannot be determined
explicitly, they can be incorporated in the estimation procedure. The distinction between
systematic and stochastic errors is a convenient way to handle the error problem but by
no means a law of nature. Stochastic errors often become systematic when an increase
in measurement accuracy makes it possible to identify an underlying non-stochastic phe-
nomenon.
The standard geodetic procedure to process measurements proceeds therefore in two steps.
First, a measurement model is developed, comprising observable plus systematic effects.
Second, a stochastic model is formed which characterizes the average behavior of the re-
maining errors and can be used in an estimation process. Representation of the Earth and
its gravity field is the next key word in the definition. Representation is the second stage in
the modeling process. Instead of modeling an individual observation to obtain the observ-
able, a set of observables is now used to determine an adequate representation of the Earth
and its gravity field. This modeling is therefore task oriented and the same observables can
be used for different tasks. Traditionally, the two major tasks of geodesy have been defined
as positioning and gravity field determination. In the same vein, observables have been
subdivided into geometrical observables, like distances and directions, and physical ob-
servables, such as gravity and its gradients. This distinction is not made anymore because
most observables can be used either for positioning or gravity field determination. If all
observables are used to determine positions and gravity field components simultaneously,
the terms integrated geodesy or operational geodesy are used.
Starting from the measurement, a measurement model is formulated that distinguishes
between the observable, the systematic errors or biases b, and the noise n.
l = L + , (1)

C Alexander Braun (2006-2010)


9 ENGO421: COORDINATE SYSTEMS
with l = measurement, L = observable, and  = errors.
 = n + b + ge (2)
with n = random errors, b = systematic errors or bias, and ge = gross errors/blunders.
Neglecting the gross errors it can be seen that these three quantities (l, L, ) cannot be sep-
arated on the basis of a single measurement. The observable can be estimated if redundant
measurements are available. The bias can either be obtained by a calibration procedure
or again by estimation. In some cases, it can also be eliminated by differencing between
observations. The random error or noise is described in a statistical manner, usually by
defining its mean and variance from previous experience. It enters into the estimation
process via a covariance matrix. Once the measurement model has been set up, the ob-
servable L is parameterized in terms of the coordinates x and the gravity potential W (will
be introduced in chapter 2). The resulting equation is nonlinear. The linearization of this
equation is done by defining approximate coordinates x and a reference potential U .

L = F (x, W ) (3)
L = f (x − x0 , W − U ) (4)

By forming the difference between the actual parameters (x, W ) and the reference model
(x, U ), by expanding into a Taylor series about (x, U ) and keeping only the first term, the
linearized model is obtained. It can take three different forms.
L = Aδx + BδW + b (5)
L − b = Aδx + BδW (6)
L − b − BδW = Aδx, (7)
with
δf (x, W )
A= (8)
δx
δf (x, W )
B= , (9)
δW
both at x0 , W0 . In the first case, the coordinate corrections δx, the gravity field correction
parameter δW , and the biases b are all estimated. This is the case of integrated geodesy.
In the second case, the bias term is either obtained by calibration and subtracted from the
observable, or is eliminated by differencing. Thus, only δx and δW have to be estimated.
In the third case, sufficiently accurate knowledge of the gravity field is available and the
corrections (reductions) to the observables can be made. In this case, only the coordinate
corrections δx have to be estimated. Depending on the model chosen, either δx only, or
δx and δW , or δx, δW , and b are estimated. The estimated δx and δW are used for the
representation of the Earth’s surface and/or its gravity field. Measurement and measure-
ment space also change in time. This is true for man-made changes which usually occur
on a time scale of a few years, as for instance subsidence in mining areas, as well as for
changes generated by geodynamic or large-scale climatic processes which occur on a scale
of ten thousand years and up, as for instance post glacial rebound (also know as GIA,
glacial isostatic adjustment), tectonics such as plate motion or the decrease of the Earth’s
rotation rate. The latter group of problems is at the centre of research as this needs to be

C Alexander Braun (2006-2010)


10 ENGO421: COORDINATE SYSTEMS
incorporated if a stable coordinate or reference systems is required. It cannot be dealt with
in detail in this introductory course but some attention will be given to the effect of these
processes throughout the course in examples and applications. It is obviously a concern
of geodesists that the coordinate system to which they refer the measurements to is stable
in time or can be reduced mathematically to a stable system at a certain epoch. If a mea-
surement is repeated after a certain period of time, it should show the same result except
for measurement errors. Changes of a local or regional scale can usually be detected by
referring all measurements to a global reference system. Changes of a global nature can
only be detected by using a reference outside the Earth, which is not part of the dynamics,
such as a set of fixed stars, or a selected group of quasars or a stable system of satellites.
The efforts to define and realize an inertial frame of reference are all part of this problem.
Which problems have to be solved when using observables for the representation of the
Earth and its gravity field? A simple example will be used to discuss some of the major
points. Let us assume that a small part of the Earth’s surface has to be represented in form
of a topographic map. Which observables are important in this case? Obviously heights,
because they are represented on a topographic map. Let us assume for a moment that we
have an instrument which allows us to measure heights wherever we want. How can it be
used to produce a topographic map? Obviously we have to know where the heights have
been measured in order to plot them on the map. This means that horizontal coordinates
are needed. They will connect the discrete height observables. It also means, that we have
found a convenient three-dimensional coordinate system for the task at hand. It consists
in this case of coordinates on a reference surface to which the heights are orthogonal. On
the reference surface, a system of horizontal coordinates can be defined in several ways. In
principle, the choice of an adequate coordinate system to connect discrete geodetic observ-
ables is always needed to represent the Earth and its gravity field. Different applications
require different coordinate systems and the term adequate is therefore task related. The
definition of appropriate local, global, and inertial coordinate systems and of the transfor-
mations between them is therefore at the centre of this course. In the discussion so far,
we have tacitly assumed that we know what a height is. Since it is a very intuitive con-
cept (height is up), we seldom stop to think how it is defined. Obviously height is related
to some reference surface and is measured as the orthogonal distance to this reference
surface. Height above a plane is an obvious example. Since a representation of height
observables in a mapping plane is needed, a simple approach would be to define height
with respect to some plane tangential to the Earth in the middle of the map sheet. This
will obviously not work too well because the Earth has a curved surface and a large lake
for instance would show height differences in this representation. Since a lake is a level
surface, such a representation would contradict intuition. In addition, this representation
would produce jumps in height when going from one map sheet to the next, an effect
which is very undesirable from a practical point of view. A reference surface is therefore
needed which represents the global shape of the Earth sufficiently well. One reference
surface which is often used is that of a globally best fitting ellipsoid of revolution. Height
is defined as the orthogonal distance with respect to this surface. Since it is a smooth
and reasonably simple surface, its mathematical description is simple, computations are
easy, and mapping is consistent. This is one reason why the determination of a best fit-
ting global ellipsoid has been a central task of geodesy for many years. Today, it can be
considered as solved with a satisfactory degree of accuracy through the use of dedicated
satellite gravity missions. For the topographic mapping problem, the choice of an ellip-
soid as reference surface for height is, however, not ideal. The ellipsoid is a mathematical
abstraction which has no physical equivalent in nature. This means that it is not possible

C Alexander Braun (2006-2010)


11 ENGO421: COORDINATE SYSTEMS
to directly physically sense the ellipsoid. It is not necessarily true that water flows from a
height of 2 m above the ellipsoid to a height of 1.8 m above the ellipsoid, as the ellipsoid
does only approximate the physical shape of the Earth. The reference surface for most
height systems is therefore defined in a physical sense. The term ”height above sea level”
indicates that. It is a level surface which in a first approximation represents the idealized
surface of the oceans and is called geoid. In land areas, this surface can only be deter-
mined from a knowledge of the gravity field of the Earth. Thus, the apparently simple task
of producing a topographic map, can in principle not be solved without a knowledge of the
Earth’s gravity field. This will be a recurring theme of the following chapters. Positioning
and gravity field determination are intertwined because gravity determines the structure
of the space in which the measurements are taken and affects the instruments, the objects
to be measured and the measurement process.

C Alexander Braun (2006-2010)


12 ENGO421: COORDINATE SYSTEMS
2 System of Natural Coordinates
Coordinate systems are fundamental in science and engineering to refer observations to
a reference which is unique and can be understood by others. Coordinates do not make
sense without a coordinate system. For instance, an engineer gives his position relative
to the class room door, the only way to understand this position is the knowledge of the
reference, here the class room door. The objective of this chapter is to establish a coor-
dinate system which can be used to refer geodetic positions to and which is physically
meaningful. In other words, the system should follow physical parameters which we know
and understand intuitively, e.g. the directions of up and down follow the direction of the
gravity vector or plumbline. In geodesy, coordinate systems are a convenient way to ex-
press general physical laws and to relate them to geodetic measurements. In principle,
the choice of such coordinate systems is arbitrary, e.g. the reference can be the class room
door in the back or in the front. The engineer or scientist will, however, make a choice
which allows to communicate the coordinates to others without complication. Also, it is
advisable to select a system with specific properties in order to simplify the representa-
tion of the measurements or the computation of results. One such system is the system
of natural coordinates. Its axes are defined by directions which are physically meaningful,
namely, the directions of the gravity vector and the spin axis of the Earth. The gravity
vector defines the up-down direction, which is the direction orthogonal to a level surface,
such as a large body of undisturbed water. The spin axis of the Earth defines the North
pole where it pierces the Earth’ surface and thus the north direction. Since many geodetic
instruments such as levels, theodolites, inertial survey systems are aligned to this frame
during the set up, it can be considered as a typical coordinate frame for geodetic measure-
ments. It is therefore often called the system of natural coordinates outlining that natural
or physical parameters define its orientation. It can also be considered as the coordinate
system describing the geometry of the gravity field as the gravity vector defines the verti-
cal axis. In order to understand this concept, Newton’s laws and specifically Newton’s law
of gravitation and its mathematical representation must be discussed. Later, the second
natural coordinate reference, the direction of the Earth’ spin axis will be introduced.

2.1 Newton’s Law of Gravitation - Gravitational Acceleration and Po-


tential
The principle of attraction of physical bodies has been mathematically formulated by New-
ton (1687) in his law of universal gravitation.

1. Every body continues in its state of rest of of uniform motion in a straight line unless
it is compelled to change that state by an external impressed force.
2. The rate of change of momentum of the body is proportional to the force impressed
and is in the same direction in which the force acts.
3. To every action there is an equal and opposite reaction.

Sir Isaac Newton (1642-1727) was the first scientist who developed a mathematical de-
scription of these laws. Before him, Johannes Kepler (1571-1630) established similar laws
for the motion of the planets and the Moon from empirical relations derived from obser-
vations. As we will later see, Kepler’s laws can be derived from Newton’s laws with certain

C Alexander Braun (2006-2010)


13 ENGO421: COORDINATE SYSTEMS
simplifications. Newton did not include relativistic effects in his laws as he was not aware
of their existence, later in the 20th century, relativistic effects have been included in Ein-
stein’s theory of relativity. For instance, the Newtonian momentum p~N of a body is the
product of its velocity ~v and mass m0 :
p~N = m0~v (10)
Once the velocity increases towards the speed of light c, the ”rest“ mass m0 becomes a
relativistic mass mRel .
m0
mRel = q 2
= γm0 (11)
1 − vc2
Newton’s momentum p~N is no longer valid and becomes the relativistic momentum p~Rel .
m0
p~Rel = q ~v (12)
v2
1− c2

It is further worth to notice that for small velocities v, the relativistic momentum becomes
Newton’s momentum. The equivalence of mass and velocity is a topic of theoretical physics
and will not be discussed here in more detail. However, an example using the equations
above results in the fact that you have to reach 14% of the speed of light, or about 42 ·
106 m/s before the mass changes by 1%. Newton’s second law states that the momentum
change wrt time is proportional to the force impressed:
~
dp
F~ = (13)
dt
In classical mechanics, the mass is considered constant, and if the momentum doesn’t
change or no force is impressed, the equation is equal to zero.
~
dp
F~ = = m0~a = 0, (14)
dt
with the acceleration ~a. In conclusion, classical mechanics and Newton’s laws are sufficient
for most geodetic applications and relativistic effects are mostly ignored with the exception
of satellites orbiting the Earth, where these effects are accounted for already, e.g. in GPS.
The previous equation leads to Newton’s first law which states that two point masses m
and m0 separated by a distance l, attract each other with a force F~ which is proportional to
the product of the two masses and inversely proportional to the square of their distance:
0
mm
F~ = G 2 (15)
l

The force F~ known as the gravitational force or gravitational attraction is directed along
the line connecting the point masses m and m0 . The constant of proportionality G, called
Newton’s gravitational constant, has the value
G = 6.6710−8 cm3 g −1 s−2 = 6.6710−20 km3 kg −1 s−2 . (16)
G is one of the least accurate physical constants known to a relative precision of 10−4
while most other physical constants are determined better than 10−7 . Consequently, G is

C Alexander Braun (2006-2010)


14 ENGO421: COORDINATE SYSTEMS
not always considered a constant, however, experiments designed to determine G could
not prove that G is not a constant. The force F~ has therefore the unit cm g s−2 = 1 dyne =
10−5 N ewton. The attraction of the masses m and m0 is completely symmetrical as stated
by Newton’s 3rd law. It is convenient, however, to consider one of them (e.g. m0 ) as the
attracted mass and the other (m) as the attracting mass. Moreover, by dividing Newton’s
law of gravitation by m0 , the attracted mass is chosen as the unity of mass. This changes
the unit to cm s−2 = dyne g −1 = 1 Gal. The unit “Gal” is named after the Italian physicist
Galileo Galilei (1564-1642) and is frequently used in geodesy and geophysics, particu-
larly to describe variations of the gravity vector at the Earth surface for exploration and
surveying. Hence, it has the unit of an acceleration, e.g. m s−2 .
1 Gal = 10−2 m/s2 , or 1 mGal = 10−5 m/s2 (17)
Assuming a Cartesian coordinate system x, y, z, the point mass m at P (x1 , y1 , z1 ) attracts
the unit mass m0 = 1 at point P 0 (x2 , y2 , z2 ) with the force F:
m ~r2 − ~r1
F = −G (18)
l2 |r2 − r1 |

Figure 1: Gravitational attraction between mass points.

Let us define the following quantities (Figure 1): the scalar distance l between the mass
∆~r
points l = |r2 − r1 | = |∆~r|, the distance vector ∆~r = ~r2 − ~r1 , and the unit vector ~e12 = |∆r| .
The transition to Cartesian coordinates is done by expressing ∆~r by
 
x2 − x 1
∆~r = ~r2 − ~r1 =  y2 − y1  . (19)
z2 − z 1

The gravitational acceleration ~b becomes:

~b = −G m ~e12 = −G m ∆~r = −G m (~r2 − ~r1 ), (20)


l2 l2 l l3
C Alexander Braun (2006-2010)
15 ENGO421: COORDINATE SYSTEMS
or in Cartesian coordinates
 
m x2 − x 1
~b = −G 3
 y2 − y 1  (21)
((x2 − x1 )2 + (x2 − x1 )2 + (x2 − x1 )2 ) 2 z2 − z 1

Figure 2: Mass Point P 0 (x2 , y2 , z2 ) attracted by the Earth.

While this relation between gravitational acceleration/attraction and mass was derived
for a mass point, it is straightforward to consider several mass points which make up an
extended body of mass by summing up the individual components (Figure 2):
n
~b = −G
X mi
3
∆~ri (22)
i=1 li

Let the mass element become infinitesimally small, so that the ratio between the mass
element and its volume at all points Q can be expressed by the density ρ and the sum
becomes an integral.
∆m
ρ = lim (23)
∆V →0 ∆V

The resulting expression for the gravitational acceleration of an extended mass body be-
comes: Z
~b = −G ρ(Q)
3
∆~r dV ol (24)
V ol li

In conclusion, if the entire mass would be concentrated in the centre of gravity, ~b would be
identical, but the body must consist of concentric spheres of constant density or the entire
sphere must be of homogeneous density.
Now, we have developed a set of equations which describe the gravitational acceleration
with a triple of Cartesian coordinates x, y, z. Hence, three scalars are required to derive
~b. The next step is to find a physical parameter which describes the gravitational field
with just one scalar. This is the search for the gravitational potential. There are two

C Alexander Braun (2006-2010)


16 ENGO421: COORDINATE SYSTEMS
ways of introducing the potential, a physical way which derives the potential from the
acceleration vector, or a mathematical way, which assumes a certain function and proves
that this function describes the potential. Here, we follow the classical geodetic way, which
assumes a function and tests if this function describes the gravitational potential in relation
to ~b. The gravitational vector field ~b is a conservative field, because to transport mass from
point A to point B, the same amount of work needs to be done, no matter what path is
chosen to move the mass through the field. This can also be expressed by stating that ~b is
curl-free.
curl ~b = rot ~b = ∇ × ~b = 0 (25)
It is also known that the curl of every gradient field equals zero, thus

rot grad V = ∇ × ∇V = 0. (26)

Herein, grad V could be ~b and V is a scalar field. Then ~b can be a gradient field of a scalar
field, here the gravitational potential V . V requires only one number while ~b requires three
numbers to describe it entirely. What function V would fulfill the relation with ~b? Let us
assume that V takes the following form:
Gm
V = (27)
l
Computing the gradient vector of the scalar function V results in:
 
  ∂V
Vx  ∂x2 
∂V
gradV = Vy  = 

 ∂y2

 (28)
Vz ∂V
∂z2

Deriving the partial derivative of V wrt to x2 .


∂V ∂ Gm ∂ 1 Gm ∂l Gm 2(x2 − x1 )) x2 − x 1
Vx = = = Gm =− =− 2 = −Gm (29)
∂x2 ∂x2 l ∂x2 l l2 ∂x2 l 2l l3
In a similar way, differentiating the scalar function V with respect to the variables y2 and
z2 gives Vy and Vz .
 
Gm  x2 − x1 
grad V = − 3 y2 − y 1 (30)
l z2 − z 1
Comparing this last expression with the equation for ~b results in
~b = grad V. (31)

The scalar function V (x, y, z) is called the gravitational potential. Its physical interpreta-
tion is given by the work needed to bring a unit mass from infinity to the point P 0 (x2 , y2 , z2 ).
This equation defines ~b as a conservative vector field. In physical terms, a vector field is
called conservative if the total energy of a body moving in this field is conserved, i.e. con-
stant. In mathematical terms it means that there exists a scalar function V such that for
each point in this field ~b is equal to the gradient of the scalar function V . Typical properties
of a conservative field are:

C Alexander Braun (2006-2010)


17 ENGO421: COORDINATE SYSTEMS
1. The integral
R P2
~bdr
~ is path independent.
P1

2.
H
~bdr
~ = 0, i.e. integration over a closed path is zero.

3. dV = ~bdr,
~ i.e. dV is an exact differential.

Property 1 states that in a conservative vector field the work done in moving the body
from point P1 to P2 is independent of the path taken. Property 2 is a simple consequence
of Property 1.
The relation between ~b and V and the fact that the gravitational field is a conservative
vector field is of basic importance. It means that the vector field described by three scalars
can be replaced by a scalar field consisting of only one scalar. The vector field can then be
obtained by differentiating the scalar field with respect to the three coordinate directions.
So far, a simple mathematical model consisting of two mass points attracting each other
has been considered. Such a model is frequently applied in celestial mechanics as a first
approximation for the solution of the two-body problem. It is possible to use this model
because the distances between the celestial bodies are in most cases so large that the
celestial bodies themselves can be considered as mass points. For measurements on the
surface of the Earth, this simple model is usually not applicable because the attracting
masses cannot be considered as mass points. In some cases, the potential can be modelled
by a system of point masses as it was assumed earlier with ~b. It is occasionally used for
local gravity field approximation. In general, however, the dimensions of the Earth have to
be taken into account as well as the density distribution in its interior. Thus, the attraction
of a mass point P by the Earth will be described as the attraction by a volume with a
continuous mass distribution. It is given that the sum of the individual contributions of
the mass elements result in V :
n
X mi
V =G (32)
i=1 li

For infinitesimal mass elements, the sum turns into an integral over the volume of the mass
and the gravitational potential V can be expressed by:
Z
ρ(x, y, z)
V =G dx dy dz (33)
V ol l
Differentiating this equation results in the equation which was previously derived for the
gravitational acceleration of an extended mass object.
Z
~b = −G ρ(x, y, z)
∆~r dx dy dz (34)
V ol l3
If the density distribution of the Earth ρ(x, y, z) is known, then both the gravitational poten-
tial and the gravitational attraction can be computed. In general, the density distribution
is not known with sufficient accuracy to use this approach. These equations are, how-
ever, fundamental for the definition of the relation between gravitation and mass density
distribution.

C Alexander Braun (2006-2010)


18 ENGO421: COORDINATE SYSTEMS
2.2 Gravity Potential and Centrifugal Potential

Up to now, only the gravitational attraction ~b and the gravitational potential V have been
discussed. The gravitational attraction is, however, not the only force acting on a body at
rest on the Earth’s surface. Due to the fact that the Earth is rotating about its axis of inertia,
an additional force, called centrifugal force, has to be considered. Its direction is always
orthogonal/normal/perpendicular to the rotation axis of the Earth. It is an apparent or
inertial force because it is completely dependent on the rotation of the Earth with respect to
an inertial frame of reference; as soon as the attracted mass stops rotating, the centrifugal
force vanishes. Assuming again that the attracted mass is equal to unity and using it as a
divisor, the total acceleration acting on a body at rest on the Earth’s surface is the resultant
of gravitation and centrifugal acceleration and is called gravity, i.e.
gravity = gravitational + centrifugal acceleration
The following will establish the equations which are required to derive the centrifugal
acceleration and later also the centrifugal potential. To explain centrifugal acceleration,
consider a simple example (Figure 3). Let the point P rotate about a fixed origin 0 at the
end of a bar which is infinitely thin and without mass. Denote the distance of the rotating
point P from the rotation centre by p~, the linear velocity by ~vl and the angular velocity by
ω. From the small angle approximation, we know that the arc segment s can be related to
the radius vector p~ and the rotation angle Θ:
s = Θp (35)
Two times differentiation wrt time results in an expression for the tangential acceleration:
~
ds dΘ
=p = ~vl = pω (36)
dt dt
~l
dv dω
=p = ~at (37)
dt dt
It shows that for ω = 0 or ω = constant, the tangential acceleration vanishes. More
important is the normal acceleration ~an . Again, we employ the sine law for small angles
and get two equations for the arc increment ∆~s.
∆~s ∆~v
= (38)
p ~vl
∆~s
= ~vl (39)
∆t
Substitution of ∆~s in one of the two equations results in an expression for ~an .
∆~v ~vl ∆t ∆~v ~v 2
= → = l = ω 2 p = ~an (40)
~vl p ∆t p

From the above, the centripetal (inwards directed) force F~cp can be derived.

F~cp = m~an = mω 2 l (41)

In case of the Earth, Figure 3 represents a plane orthogonal to the rotation axis of the
Earth, i.e. a section through the parallel of latitude φ = constant. The point P is a point

C Alexander Braun (2006-2010)


19 ENGO421: COORDINATE SYSTEMS
Figure 3: Circular motion and centrifugal acceleration.

on the Earth’s surface, a distance p away from the nearest point on the rotation axis, and
ωE is the angular velocity of the Earth, considered to be constant in this example. Note
that the rotation period of a planet and the length of day are different quantities as they
have different references, i.e. fixed stars or the Sun, respectively. Figure 4 shows the
situation, where a Cartesian coordinate system has been chosen in such a way that its z-
axis coincides with the spin axis. The centrifugal acceleration f~c at point P (x, y, z) in this
case is equal to the normal acceleration ~an and can thus be expressed as

f~c = ~an = ω 2 p~ (42)

By writing the vector p~ as


 
x
p~ =  y  (43)
0
one obtains  
ωE2 x
~
fc = ωE2 y  ,
 (44)
0
with q
p = |~
p| = x2 + y 2 . (45)

Since the centrifugal acceleration is proportional to the distance p normal to the rotation
axis, it becomes zero at the rotation poles,

f~cP ole = 0, (46)

and will reach its maximum at the equator with an equatorial radius RE ,

f~cEq = RE ωE2 . (47)

Assuming the following function for the centrifugal potential Vc ,


1 1
Vc = ωE2 p2 = ωE2 (x2 + y 2 ), (48)
2 2

C Alexander Braun (2006-2010)


20 ENGO421: COORDINATE SYSTEMS
Figure 4: The Earth’ centrifugal force.

it can be shown that the gradient of this expression equals the centrifugal acceleration f~c .
 ∂Vc   
∂x ωE2 x
gradVc = 

∂Vc
∂y

 = ωE2 y  = f~c
 (49)
∂Vc 0
∂z

As both the gravitational and centrifugal potentials are scalar and only a function of space,
both terms can be added and the sum represents the gravity potential W .

W = V + Vc (50)

Considering the previous equations for the individual potentials, W becomes,


Z
ρ(x, y, z) 1
W =G dV ol + ωE2 (x2 + y 2 ) (51)
V ol l 2
The gradient of the gravity potential W is defined as the gravity vector ~g with the compo-
nents ~b and f~c , Z
∆~r
~g = ~b + f~c = grad W = −G ρ dV ol + ωE2 p~ (52)
V ol l 3

Note that gravitation decreases with the squared distance from the attracting masses while
the centrifugal acceleration increases with distance p from the rotation axis. The gravita-
tional vector ~b points inward while the centrifugal vector points outward. Figure 5 depicts
this situation in graphical form. The magnitude of ~g is called gravity and is measured in

C Alexander Braun (2006-2010)


21 ENGO421: COORDINATE SYSTEMS
Gal or more commonly in mGal. Its value at the poles is about 983 Gal and it decreases
systematically to about 978 Gal at the equator. This is due to differences of centrifugal
and gravitational acceleration at the equator and at the poles. The centrifugal acceleration
is about 3.4 Gal at the equator pointing outward and zero at the poles. The magnitude
of gravity is therefore reduced at the equator. The flattening of the Earth at the poles
decreases the distance to the centre of mass of the Earth by about 22 km, which is about
3%o of the Earth’s radius. This, in turn, would increase the magnitude of the gravitational
attraction ~b at the poles.

Figure 5: Interaction of gravitational and centrifugal accelerations ~b and f~c .

The combined effect of the change in centrifugal acceleration and the change in gravita-
tional acceleration due to the flattening of the Earth results in the gravity difference of
about 5 Gal between the equator and the poles. 35% are due to flattening, 65% are due
to Earth rotation. If this change was completely systematic and symmetric, a simple global
model for the change of gravity could be derived. However, due to the inhomogeneous
density distribution and related mass irregularities in the interior of the Earth, the actual
global gravity model is much more complicated. The simplified model is often used as a
first approximation and is then called the normal gravity model.

2.3 Level Surfaces and Plumb Lines


So far, only the magnitude of the gravity vector ~g has been considered. In this section, the
direction of the gravity vector and the characteristics of the surface orthogonal to it will be
discussed. Remember, the objective of this chapter is to introduce a physically meaningful
reference system. Let us start by setting

W (x, y, z) = WP = constant. (53)

The surfaces defined in this way are surfaces of constant potential, called equipotential
surfaces or in the case of the gravity potential, level surfaces. The levelling bubble of a
theodolite orients itself to lie in this level surface. Equipotential surfaces coincide with the

C Alexander Braun (2006-2010)


22 ENGO421: COORDINATE SYSTEMS
surface of a homogeneous fluid in equilibrium with no external forces except the gravity
field, which explains the term level surface. In a first approximation, the idealized surfaces
of lakes can be considered as such level surfaces. They approximate W = constant for a
specific value WP . Differentiating W = W (x, y, z) with respect to (x, y, z) gives
∂W ∂W ∂W ~
dW = dx + dy + dz = gradW dr, (54)
∂x ∂y ∂z

where dr~ T = (dx, dy, dz) is the displacement vector. Let dr


~ lie in the equipotential surface
W (x, y, z) = WP , then dW = 0 and
~ = 0 = ~g · dr
gradW · dr ~ (55)

on WP . If the dot product of two non-zero vectors (both ~g and dr~ are non-zero) is equal
to zero, then the vectors are orthogonal to each other, so that the gravity vector must be
~ In addition, this means that the gravity vector is normal to the equipo-
orthogonal to dr.
tential surface passing through the point P . It is therefore simple to find the direction
of the gravity vector on the surface of the Earth. It is orthogonal to the surface estab-
lished by a level bubble, or, in other words, the bubble represents the level surface in that
specific point. This fundamental principle is used extensively in the levelling of geodetic
instruments.

Figure 6: Level surfaces, plumblines, geoid, orthometric height H.

The lines which intersect all level surfaces of the Earth orthogonally are called plumb
lines. They are curved lines and the gravity vector is obviously tangent to the plumb line
at the points of intersection. A good approximation of such a tangent, and therefore of
the direction of gravity, is the string holding a plumb bob. Each specific WP = constant
defines a different equipotential surface (Figure 6). The particular equipotential surface

C Alexander Braun (2006-2010)


23 ENGO421: COORDINATE SYSTEMS
which coincides with the idealized surface of the oceans is called the geoid. It is assumed
that there are no forces acting on the ocean such as ocean dynamics, tides, wind, waves etc.
Then, the approximation of the mean sea level can be used to describe the geoid, however,
this is only a first order approximation and when considering the effects mentioned above,
the mean sea level is not equal to the geoid surface. The difference between them is also
referred to as sea surface topography. The name geoid was proposed by Listing to describe
the figure of the Earth. The geoid is used as a reference surface for the orthometric height
system which will be discussed later in this chapter. It defines the height H of a point at
the physical surface of the Earth by its distance from the geoid measured along the plumb
line. The following properties of the Earth’s equipotential surfaces are of importance in
geodesy.
Equipotential surfaces

• are continuous surfaces,


• never cross each other,
• are not necessarily parallel to one another,
• change their curvature smoothly from point to point.

2.4 Natural Coordinates


The objective of this chapter is to develop a system of natural coordinates using axes
defined by directions which are physically meaningful in the terrestrial space, e.g. the
directions of the gravity vector and the spin axis of the Earth. How can these spatial
directions be related to positions in an Earth-fixed coordinate system? It is obvious from
the preceding sections that the gravity potential and its gradients are important in this
context. They define the direction of the gravity vector by gradients of W in an Earth-fixed
Cartesian coordinate system x, y, z. The relationship between grad W and x, y, z will be
briefly discussed in this section. Natural coordinates will be used to define the three axes
of orthogonal coordinate systems; together with the origin, this defines a reference frame.
The simplest representation of the gravity vector is obtained in the Local Astronomic Frame
(LA) which is defined by:

• Local Astronomic Frame (LA)


• Origin: At observers point P
• Primary axis (z): Orthogonal to level surface at P (WP = constant)
• Secondary axis (x): Tangent to astronomic meridian pointing north
• Tertiary axis (y): Orthogonal to complete a left-handed system

The LA is the frame which is used to take measurements in the field. The word “local”
indicates that the frame is used in the local measurement environment (Figure 7). The

C Alexander Braun (2006-2010)


24 ENGO421: COORDINATE SYSTEMS
word “astronomic ” indicates that the natural coordinates are used which have a physical
meaning. In this system, the gravity vector has the coordinates
 
0
~gLA = grad WLA = 0  (56)
−g

To relate this representation to an Earth-fixed Cartesian system, the Conventional Terres-


trial Frame (CT) is defined:

• Conventional Terrestrial Frame (CT)


• Origin: Earth’ centre of mass
• Primary axis (z): Conventional (or mean) spin axis of the Earth
• Secondary axis (x): Intersection of the conventional (or mean) equator plane and
the mean Greenwich meridian plane
• Tertiary axis (y): Orthogonal to complete a right-handed system

This coordinate frame is of fundamental importance in geodesy. It will be more rigorously


defined in ENGO423: Geodesy, where the terms mean or conventional get a proper defini-
tion based on Earth rotation, precession and nutation. The word “conventional” indicates
that there is a definition involved, a convention or average position of the physical quanti-
ties. The word “terrestrial” describes the origin as the Earth’ centre of mass as opposed to
“local”, where the local point is the origin. The plane orthogonal to the conventional/mean
spin axis is called conventional equator plane. The direction of ~g in the CT-frame is given
by two angles, the astronomic latitude Φ and the astronomic longitude Λ. The definition
of Λ is tied to the definition of the astronomic meridian plane, i.e the mean Greenwich
meridian plane.
The astronomic meridian plane of a point P is the plane containing the gravity vector at
P and the parallel to the conventional rotation axis of the Earth through P . Thus, it is
orthogonal to the conventional equator plane. The astronomic longitude Λ is the angle
between the astronomic meridian planes of two points. The convention is that the angle
Λ is counted counterclockwise from the mean astronomic meridian plane of Greenwich.
The astronomic latitude Φ of a point P is the smallest angle between the conventional
equator plane and the vector normal to the level surface in P measured in the meridian
plane of P . The normal vector is opposite in direction to the gravity vector in P . The angle
Φ is conventionally counted from the mean equator plane positive towards the north pole
and negative towards the south pole. Since each point on the surface of the geoid can be
expressed by a pair of coordinates Φ,Λ, these coordinates can be mapped onto the surface
of a unit sphere. By connecting a specific coordinate pair to the centre of the unit sphere,
we obtain a unique spatial unit vector ~n for each pair. This unit vector gives the direction
of the vector normal to the geoid at P expressed as a function of Φ and Λ. This direction
is shown in Figure 8.
The vector p~ in Figure 8 is the projection of the normal vector ~n onto the conventional
equator plane. The angle between the vectors ~n and p~ is the astronomic latitude Φ and the

C Alexander Braun (2006-2010)


25 ENGO421: COORDINATE SYSTEMS
Figure 7: Local astronomic frame and astronomic coordinates.

angle between p~ and the mean Greenwich meridian plane is the astronomic longitude Λ.
Since |~n| = 1 and |~
p| = cosΦ, the vector p~ has the coordinates
 
cosΦ cosΛ
p~ =  cosΦ sinΛ  (57)
0

Based on this, the vector ~n can be expressed because it only differs from p~ in the z-
component.  
cosΦ cosΛ
~nCT =  cosΦ sinΛ  (58)
sinΦ sinΦ
Using this definition of the normal vector ~n, the gravity vector ~g can be expressed as

~g CT = −g ~n (59)

Finally, the gravity potential can be substituted and we obtain,


 ∂W     
∂x Wx cosΦ cosΛ
∂W
~g CT = gradW =   =  Wy  =  cosΦ sinΛ  .
 
∂y (60)
∂W Wz sinΦ sinΦ
∂z

C Alexander Braun (2006-2010)


26 ENGO421: COORDINATE SYSTEMS
Figure 8: Direction of normal vector in terms of Φ and Λ.

The last equation shows an interesting connection between physics and geometry. Starting
from physics (gravity potential), the astronomic coordinates Φ, Λ can be derived and the
geometry (e.g. the curvature of the Earth) can be determined. It defines the gradients of
the gravity potential in terms of astronomic coordinates. The reverse formulas expressing
Φ and Λ as gradients of the gravity potential, can also be obtained.
Wx2 + Wy2 = g 2 cos2 Φ (cos2 Λ + sin2 Λ) = g 2 cos2 Φ (61)
Wz −g sinΦ
q = = −tanΦ (62)
Wx2 + Wy2 g cosΦ
Wy
= tanΛ (63)
Wx
Solving for the astronomical latitude and longitude, a function of the potential gradients
can be derived.
−Wz
Φ = arctan q (64)
Wx2 + Wy2
Wy
Λ = arctan (65)
Wx
This shows that, if the gravity potential W (x, y, z) is given, the coordinates Φ and Λ can
always be determined. To describe the position of a point in three-dimensional space,
three coordinates are needed. They can be Cartesian x, y, z, curvilinear Φ, Λ, H, or any
other coordinate triple. It has been shown that Φ and Λ give the position of a point on
an equipotential surface. It makes sense, therefore, to define the third coordinate as being
orthogonal to this surface. It has been mentioned before that this coordinate is called the
orthometric height H if the reference surface used is the geoid (Figure 9). To define H

C Alexander Braun (2006-2010)


27 ENGO421: COORDINATE SYSTEMS
mathematically, let us use the equation again, which relates a displacement vector with ~g
and dW .
~
dW = ~g · dr (66)
Earlier, the displacement vector dr~ lay in the equipotential surface, and we have shown
~ point upward along the plumb line, i.e. |dr|
that ~g is normal on W . This time, let dr ~ = dH.

Figure 9: Definition of the orthometric height H.

~ then becomes 180 deg, this yields


As the angle between ~g and dr
~ = |~g | · |dr|cos(~
dW = ~g · dr ~ ~ = g dH cos(180) = −g dH
g , dr) (67)
This is the fundamental equation for the definition of heights. It relates a potential dif-
ference to a height difference. The proportionality factor is the magnitude of the gravity
vector. It can also be rewritten in the form
dW
dH = − (68)
g

The equation defines the height in terms of potential differences and gravity. Since gravity
cannot be easily measured inside the Earth, different approximations for g inside the Earth
must be made. This leads to different height systems and definitions. A height would
only be accurately determined, if we would know g along the entire plumbline between
the point P at the Earth surface and the geoid. Different height systems solve this prob-
lem in different ways, but always use an approximation of g, sometimes obtained using
measurements at the surface or sometimes by assuming normal gravity γ.
Geopotential Numbers: A height difference can be derived by knowing the potential
difference dW between two points and g. Let us define two potential values at the surface
WP and on the geoid W0 .
−dW = W0 − WP = C (69)
Herein, C is called “geopotential number”. The height can now be defined as:
C
Height = (70)

In this case, g̃ is an approximate magnitude of g derived from gP and g0 , but depending
on the choice of the approximation, different heights can be derived, e.g. orthometric

C Alexander Braun (2006-2010)


28 ENGO421: COORDINATE SYSTEMS
heights, dynamic heights, normal heights or Helmert heights can be defined. Details about
the different height systems will be discussed in ENGO423-Geodesy. Geopotential num-
bers are expressed in terms of geopotential units (g.p.u.), i.e. 1 g.p.u. = 1 kGal meter. As
a rule of thumb, C in g.p.u is always close to the height above the geoid in meters, as g is
approximately 0.98 kGal. The system of natural coordinates can therefore be expressed by
either one of the two coordinate triples Φ, Λ, H or Φ, Λ, C. The first has a simple geomet-
rical explanation but introduces some assumptions in the definition of H. The second is
more precise in terms of defining the coordinates but lacks the intuitive geometrical mean-
ing. In both cases, the coordinates can be uniquely expressed by potential differences and
gradients of W , thus explaining Bruns (1878) concise statement: ”The task of geodesy
is the determination of the potential function W(x,y,z)”. It is important to note that the
determination of heights always involves the gravity field of the Earth as the gravity field
affects the measurement device, e.g. a theodolite. Once the observer changes location, the
value of ~g may change and consequently the height changes, and the levelling bubble will
be in a different orientation. This effect leads to the misclosure of leveling loops, if grav-
ity is not accounted for. A closed loop of geometric leveling leads to the misclosure and
depends on the path. The misclosure disappears once geopotential differences are taken
into account. This statement indicates that gravimetry is mandatory for high precision sur-
veying, however, the knowledge of the potential gradient in the surveyed area allows for
an estimation of the effect on the height determination. In area of small geoid gradients,
e.g. the Prairies, gravimetry is not required for most surveying applications. However, in
mountainous regions, where gravity changes significantly, it becomes important and must
be considered in order to achieve high accuracy.
As the geoid is not known perfectly, geodesists have developed alternative surfaces to
be used as a height reference. One example is the ellipsoid, or rotationally symmetric
ellipsoid. The task is to approximate the gravity field of the Earth or the equipotential
surface called geoid with a mathematical surface which can be derived exactly at every
location. In a first step, this surface could be a sphere, but the height differences would
be up to 11 km at the poles and the equator, as the Earth flattening causes a difference in
radius of about 22 km between the poles and the equator. The next and significantly better
approximation is an ellipsoid with semi-major axis a and semi-minor axis b. Once this fit
is made, the differences between the ellipsoid and the geoid are less than 100 meters. The
ellipsoid will be discussed in more detail in chapter 5. As terrestrial coordinate frames
have been defined, the definition of celestial coordinate frames is the next step which
allows also to observe the motion of the Earth from a non-terrestrial point of view.

C Alexander Braun (2006-2010)


29 ENGO421: COORDINATE SYSTEMS
3 Celestial Coordinate Systems
In the previous chapters, the coordinate systems were fixed to the Earth, which implies that
the coordinate systems move as the Earth moves with respect to other planetary objects
or the Sun. In order to identify coordinate systems, which allow us to observe the Earth’
motion, celestial coordinate systems will be discussed in this chapter. In other words, we
search for a connection between local, terrestrial, and inertial systems which allow us to
relate coordinates across different coordinate systems. In addition, the definition of the
natural system of coordinates (Φ, Λ, H) is tied to the knowledge of the potential function
W and its gradients. This function is not known with sufficient accuracy on Earth to actu-
ally determine positions on the surface by using formulas such as (64, 65). It is possible,
however, to establish the direction of the local gravity vector with good accuracy at each
point on the Earth’s surface. Is it possible to use this direction in space to determine Φ and
Λ? The answer is yes, if is direction can be related by measurements to the conventional
equator and the mean Greenwich meridian which are the reference for Φ and Λ, respec-
tively. Previous to the GPS era, the classical way of doing this is by way of astronomical
observations. (Note: Knowledge in Astronomical observations is still mandatory as the
availability of GPS is not guaranteed and the accreditation of our geomatics program re-
quires this.) This means that an operational inertial frame of reference must be implicitly
defined by making certain assumptions about the average direction of the spin axis and
the average position of the Greenwich observatory. It also means that Earth rotation with
respect to celestial objects has to be known with high accuracy. This chapter therefore
discusses celestial coordinate systems used in observational astronomy and their relation
to the terrestrial systems, assuming Earth rotation as known.

3.1 The Celestial Sphere


The celestial sphere is based on a simple concept frequently used in geodetic astronomy.
The celestial sphere is a 2-dimensional system where a point is fully described by 2 coor-
dinates only. Consider that all stars are projected onto a sphere with an infinite radius and
the sphere’s centre being the Earth’s centre. Further assume that the Earth’s movement
about the Sun can be neglected. The direction to each star is then given by the direction of
the unit vector from the Earth’s centre of mass. It pierces the celestial sphere at a unique
point which can be denoted by only two spherical coordinates. If these coordinates are
called α and δ, the unit vector can be written in the general form
 
cosδ cosα
e =  cosδ sinα  . (71)
sinδ

Since the length of each vector is infinite, it does not give the position of the star in 3D
space but only its direction from the Earth’s centre of mass. Figure 10 shows the celestial
sphere in equatorial orientation from the viewpoint of the Earth. The Earth’s spin axis
pierces the celestial sphere at the North Celestial Pole (NCP) and the antipodal South
Celestial Pole (SCP). The plane perpendicular to the Earth’s spin axis and containing the
centre of the celestial sphere is the celestial equator. A great circle through the star and
the celestial poles is called an hour circle. Each hour circle is orthogonal to the celestial
equator. A small circle parallel to the celestial equator, is called celestial parallel. The

C Alexander Braun (2006-2010)


30 ENGO421: COORDINATE SYSTEMS
plane containing the celestial poles and the zenith is called celestial meridian or observer’s
hour circle.

Figure 10: Top: Celestial Sphere from a Terrestrial Point of View. Bottom: Celestial Sphere
from the Viewpoint of the Sun-Earth/Moon System

C Alexander Braun (2006-2010)


31 ENGO421: COORDINATE SYSTEMS
Figure 10 shows the celestial sphere in an ecliptic orientation, i.e. from the viewpoint
of the Sun-Earth/Moon system. The Earth/Moon system, or in fact its barycentre (the
barycentre is the centre of mass of two bodies), orbits the Sun in a plane which is inclined
with respect to the celestial equator plane by an angle  = 23.5o . The orbital plane is called
the plane of the ecliptic and can also be considered as the apparent path of the Sun as seen
from the Earth/Moon barycentre. The ecliptic is defined by two vectors. The first one is the
vector between the centre of the Sun and the barycentre of Earth/Moon system, the other
one is the inertial velocity vector of the barycentre orbiting about the Sun. The ecliptic
defined in this way does not contain perturbations from the planets of the solar system,
where Venus and Jupiter cause the largest effects, however, they are small and within 2”
of the Sun’s apparent path. The northern ecliptic pole (NEP) and the southern ecliptic pole
(SEP) are defined by the points where a normal to the ecliptic plane and through the sphere
centre pierces the celestial sphere. The vernal equinox is the point where the apparent path
of the Sun crosses the celestial equator in spring coming from the southern and moving to
the northern hemisphere. In fall, the apparent path of the Sun crosses the celestial equator
again, but in this case passing from the northern to the southern hemisphere; this point is
known as the autumnal equinox. Vernal and autumnal points are thus the points on the
celestial sphere which define the intersection of the ecliptic with the celestial equator. They
are used to define the direction of one coordinate axis in most celestial coordinate systems.
Summer and winter solstices are points on the ecliptic which have an angular separation
of 90 from either equinox. They mark the largest distance of the Earth’s orbit (or Sun’s
apparent path) from the celestial equator. The equinoctial (solstitial) colure is the great
circle through the celestial poles and the equinoxes (solstices) or, in other words, the hour
circles of the equinoxes (solstices). Equinoctial and solstitial colures are orthogonal to each
other. Further, we define the nadir as the point on the celestial sphere, where the extension
of the local gravity vector pierces the sphere and we define the zenith antipodal, where the
extension of the gravity vector in the opposite direction pierces the sphere. The zenith can
be generally described as directly above the observer, with reference to the gravity vector
in the observers point. The following list summarizes the astronomical quantities used in
this chapter:

• Celestial poles NCP, SCP: Earth spin axis extension piercing celestial sphere
• Celestial equator: extension of Earth equator plane to celestial sphere
• Hour circle: great circle with NCP, SCP and normal to celestial equator
• Celestial parallel: Any circle parallel to celestial equator plane
• Observer related quantities:
• Zenith (Z): Direction of −~g intersecting celestial sphere
• Nadir (N): Antipodal to Zenith on celestial sphere
• Celestial horizon: Great circle orthogonal to Z, N
• Celestial vertical circle: Greta circle with Z, N, and ~g and orthogonal to celestial
horizon
• Celestial meridian: Great circle with NCP, SCP, and Z
• Observer’s hour circle: identical with celestial meridian

C Alexander Braun (2006-2010)


32 ENGO421: COORDINATE SYSTEMS
• Prime vertical: Celestial vertical circle orthogonal to celestial meridian
• Ecliptic: Plane stretched out by the apparent path of the Sun about Earth/Moon
barycentre
• North/South ecliptic poles (NEP, SEP): Axis pierces celestial sphere and goes through
centre of sphere
• Angel of the ecliptic: 23.5o angle between celestial equator and ecliptic
• Vernal equinox: Point where ecliptic intersects celestial equator from the South
• Autumnal equinox: Point where ecliptic intersects celestial equator from the North
• Winter solstice: 90o separation from equinox on ecliptic, Southern hemisphere wrt
celestial equator
• Summer solstice: 90o separation from equinox on ecliptic, Northern hemisphere wrt
celestial equator
• Equinoctial colure: Hour circle with both equinoxes
• Solstitial colure: Hour circle with both solstices and orthogonal to equinoctial colure

In order to imagine the three different types of reference systems, consider that there are
three equatorial planes and three types of poles orthogonal to these planes. The reason for
that is that these systems have a different point-of-view or purpose. While we know the
tilt between the ecliptic and the celestial equator, the tilt wrt the celestial horizon changes
with the observers location and depends on the changing gravity vector direction. Hence,
there is no constant geometric relation between the celestial horizon and the other two
equatorial planes.

1. Observer related: Z and N as poles, celestial horizon as equatorial plane


2. Terrestrial: NCP and SCP as poles, celestial equator as equatorial plane
3. Sun-Earth/Moon: NEP, SEP as poles, ecliptic as equatorial plane

In order to define the following four celestial coordinate systems (H, RA, HA, E), three
basic assumptions will be made:

• Stars have fixed positions on the celestial sphere


• The celestial sphere has an infinite radius
• The celestial sphere has its origin at the Earth’s centre of mass

These assumptions simplify the definition of celestial coordinate systems because i) star
positions can be used as a uniform reference system for all observations from the Earth,
and ii) star positions can be cataloged by using two coordinates only. Note that the as-
sumption about the geocentric coordinate system is specific to geodetic astronomy, while
in most other cases, heliocentric systems are used with the Sun in the celestial sphere’s
centre. The above assumptions give a useful model of physical reality. The corrections to

C Alexander Braun (2006-2010)


33 ENGO421: COORDINATE SYSTEMS
the coordinates due to the movement of the spin axis do not exceed 50” per year. The
corrections for the centre/origin shift from heliocentre to observer’s position (topocentric)
are below 1”. All corrections are time dependent and can be computed from parameters
given in star catalogs. In accordance with the third assumption, all celestial coordinates
systems discussed here have a geocentric origin. In each celestial coordinate system, the
positions are fixed by two spherical coordinates: the first is defined in the primary plane,
the second in the secondary plane. Depending on the definition of the primary plane and
the purpose of the coordinate systems, we distinguish the following celestial coordinate
systems:

• Horizon System (H)


• Right Ascension System (RA)
• Hour Angle System (HA)
• Ecliptic System (E).

The different systems will be defined by the direction of their primary, secondary and
tertiary axes. Thus, the primary axis defines the primary plane, the secondary axis the
secondary plane, etc. In all figures, S indicates the position of the star on the celestial
sphere.

3.2 The Horizon System - H


The horizon system is a typical observational system, it is comparable to the local astro-
nomical frame (LA) as the observers gravity vector defines the primary axis/plane. Its
coordinates are the azimuth A and zenith distance z or the altitude angle a, which is the
corresponding angle to z, e.g. a = 90o − z. a, z correspond directly to the readings of
a leveled theodolite. The horizon system is an earth-fixed system and thus rotates with
respect to an inertial frame of reference.

• Horizon System (H)


• Origin: At observer’s position P (topocentric in geodetic astronomy)
• Primary axis (z): Zenith at P (normal toWp = constant)
• Secondary axis (x): Tangent to the celestial meridian at P (Z - NCP) pointing north
• Tertiary axis (y): Pointing east in the celestial horizon plane to complete a left-
handed system (LHS)

From Figure 11 we can derive, with pH being the horizontal component and z H the vertical
component of the distance vector

pH = sinz = cosa (72)


z H = cosz = sina (73)

C Alexander Braun (2006-2010)


34 ENGO421: COORDINATE SYSTEMS
Figure 11: The Horizon system - H

and, since

x = p cosA (74)
y = p sinA (75)

the unit vector becomes

x H
     
sinz cosA cosa cosA
eH =  y  =  sinz sinA  =  cosa sinA  . (76)
z cosz sina

3.3 The Right Ascension System - RA


The Right Ascension System (RA) is a close approximation of an inertial frame of reference
(Figure 12). It is thus “fixed” with respect to distant galaxies and its two coordinates, right
ascension α and declination δare used to catalog stars. It is also used as a reference
in satellite geodesy providing the inertial frame of reference for Newton’s equations of
motion. The star catalog can be used in positioning and is updated regularly with new
observations of the two coordinates α and δ, which fully describe the position of a star on
the celestial sphere. the right ascension angle is measured in easterly direction from the
x-axis and has the units of 0-24 hours.

C Alexander Braun (2006-2010)


35 ENGO421: COORDINATE SYSTEMS
• Right Ascension System (RA)
• Origin: Earth’s centre of mass
• Primary axis (z): Spin axis of the Earth, NCP, SCP
• Secondary axis (x): Direction to vernal equinox
• Tertiary axis (y): Completing a right-handed system (RHS)

Using z = sinδ and p = cosδ, the unit vector takes the form
 RA  
x cosδ cosα
eRA = y 
 = cosδ sinα  .
 (77)
z sinδ

Figure 12: The Right Ascension System - RA

3.4 The Hour Angle System - HA


The hour angle system is a transformational system and connects the observational system
(H) to the catalog system (RA) (Figure 13). Its coordinates are the hour angle h and the

C Alexander Braun (2006-2010)


36 ENGO421: COORDINATE SYSTEMS
declination δ. The hour angle is defined as Sidereal time (ST) of the vernal equinox minus
the right ascension α of the star. Thus, time is one of the variables defining this coordinate
system. Note the equivalence of time and horizontal angles. The hour angle is measured
in westerly direction in units of 0-24 hours.

• Hour Angle System (HA)


• Origin: Earth’s centre of mass
• Primary axis (z): Spin axis of the Earth, NCP, SCP
• Secondary axis (x): Intersection of celestial meridian/observer’s hour circle and ce-
lestial equator plane
• Tertiary axis (y): Completing LHS

The unit vector takes the form


 HA  
x cosδ cosh
eHA = y  =  cosδ sinh  . (78)
z sinδ

Figure 13: The Hour Angle System - HA

C Alexander Braun (2006-2010)


37 ENGO421: COORDINATE SYSTEMS
3.5 The Ecliptic System - E
The ecliptic system is considered the most stable reference system and thus represents the
best realization of an inertial frame of reference (no acceleration) (Figure 14). It is used
particularly for Sun catalogs. Its coordinates are ecliptic longitude λ and ecliptic latitude
β, and both are referred to the ecliptic plane.

• Ecliptic System (E)


• Origin: Earth’s centre of mass
• Primary axis (z): Orthogonal to the ecliptic plane (apparent path of the Sun) and
through the centre of the celestial sphere
• Secondary axis (x): Pointing towards vernal equinox
• Tertiary axis (y): Completing a RHS

The unit vector takes the form


 E  
x cosβ cosλ
eE =  y  =  cosβ sinλ  . (79)
z sinβ

Figure 14: The Ecliptic System - E

C Alexander Braun (2006-2010)


38 ENGO421: COORDINATE SYSTEMS
3.6 Reflection and Rotation Matrices
In order to transform between different coordinate systems, the required mathematical
operations are the reflection and the rotation of coordinate axes. The tools to perform
these operations can be generally expressed by reflection and rotation matrices. Consider
the following example: Two coordinate systems (x, y) and (x0 , y 0 ) have the same origin,
but their orientation in space is different (see Figure 15). In order to express coordinates
obtained in one system by coordinates in the other system, let us write down the relation
between the coordinate pairs and the angles µ and ν in Figure 15.

Figure 15: Rotation of a Coordinate System

x0 = s cos(µ − ν) (80)
y 0 = s sin(µ − ν) (81)
x = s cosµ (82)
y = s sinµ (83)
By solving the argument of the sin and cos term and substituting s cosµ by x, and s sinµ
by y,
x0 = s cosµ cosν + s sinµ sinν = x cosν + y sinν (84)
y 0 = s sinµ cosν − s cosµ sinν = y cosν − x sinν, (85)
or in matrix notation,
    
x0 cosν sinν x
= . (86)
y0 −sinν cosν y
While the above rotation was only in 2-d space, it is straightforward to realize that the
rotation actually was performed about a third axis z which is orthogonal to x and y. the

C Alexander Braun (2006-2010)


39 ENGO421: COORDINATE SYSTEMS
third component can thus be introduced to the matrix.
      
x0 cosν sinν 0 x x
 y 0  =  −sinν cosν 0   y  = R3 (ν)  y  (87)
z0 0 0 1 z z
The rotation matrix R3 (ν) describes a rotation by the angle ν about the z-axis. In an analog
way, the rotation matrices for rotations about the other two axis can be derived.
 
1 0 0
R1 (ν) =  0 cosν sinν  (88)
0 −sinν cosν
 
cosν 0 −sinν
R2 (ν) =  0 1 0  (89)
sinν 0 cosν

Reflection matrices are even simpler. Applying a reflection matrix to a coordinate system
or coordinates results in changing the direction of one axis, or simply a sign change for
one component of the coordinates. Reflection matrices are denoted by the axis which is
reflected, e.g. P2 reflects the y-axis. The three refection matrices are as follows:
 
−1 0 0
P1 =  0 1 0  (90)
0 0 1
 
1 0 0
P2 =  0 −1 0  (91)
0 0 1
 
1 0 0
P3 =  0 1 0  (92)
0 0 −1
A combination of rotation and reflection matrix operations is often required to transform
from one coordinate system into another. It is therefore useful to investigate the properties
of these operators. The product of a reflection and rotation matrix is commutative, if their
index is identical.
Pi Ri (ν) = Ri (ν) Pi , f or i = 1, 2, 3 (93)
For different indices, the sign of the argument changes.
Pi Rj (ν) = Rj (−ν) Pi = Rj−1 (ν) Pi = RjT (ν) Pi (94)

These mathematical tools will be extensively used in the next section for transforming
celestial coordinate systems, but also in chapter 6 - Coordinate transformations.

3.7 Transformations Between Celestial Systems


Transformation between HA and RA

The HA and RA systems are shown in Figure 16. Both systems have the same primary
plane, the celestial equator, and also share the declination δ as one coordinate. Their

C Alexander Braun (2006-2010)


40 ENGO421: COORDINATE SYSTEMS
coordinates in the equatorial plane are related by
ST = α + h, (95)
where ST is the hour angle of the vernal equinox which is also known as sidereal time.
However, the two systems differ in their handedness. The transformation consists of two
steps:

• Left-handed into right-handed system: apply P2


• Rotation about z-axis by −ST : apply R3 (−ST ) = R3 (−α − h)

Expressed through a unit vector, the transformation can be written as:


eRA = R3 (−ST ) P2 eHA (96)

Figure 16: Hour Angle System transformation to the Right Ascension System - HA-RA

It is worth to realize that ST models earth rotation and therefore connects earth-fixed
terrestrial coordinate systems to celestial coordinate systems. Definitions of sidereal time
vary with the definitions of the vernal equinox and the hour angle as we will see in chapter
4. If the true vernal equinox is used, ST becomes Apparent Sidereal Time (AST). If the
mean vernal equinox is used, ST becomes Mean Sidereal Time (MST). If instead of the
hour angle of the local meridian the hour angle of Greenwich or Greenwich meridian is
used, ST becomes Greenwich Apparent Sidereal Time (GAST) or Greenwich Mean Sidereal
Time (GMST), respectively.Especially GAST will be frequently used to describe the rotation
between celestial and terrestrial coordinate systems.

C Alexander Braun (2006-2010)


41 ENGO421: COORDINATE SYSTEMS
Transformation between RA and E

Both systems are right-handed and their x-axes coincide, both point towards the vernal
equinox. The angle between the primary planes (celestial equator and ecliptic) is the
obliquity of the ecliptic . The transformation consists of a counterclockwise rotation about
the x-axis by the angle , i.e. R1 (). Thus, the unit vectors are related through
eE = R1 () eRA (97)

Transformation between H and HA

Both systems are left-handed and have the same secondary plane (observer’s celestial
meridian). The x-axes and y-axes of both systems are pointing in opposite directions. The
primary planes (celestial horizon and celestial equator) form the angle (π/2 − Φ) where Φ
is the astronomic latitude of the observer. The situation is shown in Figure 17.

Figure 17: Horizon System Transformation to the Hour Angle System - H-HA

The transformation consists of two steps:


• Rotation about z H -axis by π: apply R3 (π)
• Rotation about new y H -axis by (Φ − π/2): apply R2 (Φ − π/2)
The unit vectors are related through:
eHA = R2 (Φ − π/2) R3 (π) eH (98)

C Alexander Braun (2006-2010)


42 ENGO421: COORDINATE SYSTEMS
Other transformations between H, RA, HA, and E

Other transformations can be designed by using the previous examples and add the opera-
tors. One example would be the transformation from the Horizon system (H) to the Right
Ascension system (RA). It would consist of the following operations:

eRA = R3 (−ST ) P2 R2 (Φ − π/2) R3 (π) eH (99)

It is important to consider the rules for the inversion of the order of rotation and reflec-
tion matrices. We will later see that another parameter, GAST, will become important for
transformations between celestial and terrestrial systems. GAST will then be treated as an
angle in the rotation matrix.

3.8 Astronomical Triangle and Spherical Trigonometry


The astronomical triangle is a concept used to relate different celestial coordinate systems
and celestial parameters. It is closely connected to spherical trigonometry. The triangle is
set up between three different points, the zenith Z, the Star, and the N CP . As the celestial
sphere is a sphere, we can use spherical trigonometry to develop relations between the
astronomical parameters, here, the sides of the triangle, which are angles. Figure 18
shows the astronomical triangle and the associated parameters. Two cases are considered,
the star is east of the zenith or the star is west of the zenith. Besides the three points
defining the triangle, there are three interior angles which are at the three points, and
three exterior angles, which represent the length of the sides. As this is on a sphere, the
sides are in fact angles as well. Figure 18 shows the two cases.

• Case 1: Star east of the zenith


• interior angles are:
hour angle: 24-h
azimuth: A
parallactic angle: ρ, which is defined by the hour circle and the vertical circle
• sides of the triangle are:
NCP to zenith: 90o − Φ
NCP to star: 90o − δ
zenith to star: z = 90o − a
• Case 2: Star west of zenith
• interior angles are:
hour angle: h
azimuth: 360o − A
parallactic angle: ρ
• sides of the triangle remain as in case 1.

C Alexander Braun (2006-2010)


43 ENGO421: COORDINATE SYSTEMS
Figure 18: Top: Astronomical triangle and astronomical angles. Bottom left: Astronomical
triangle if the star is east of the zenith. Bottom right: Astronomical triangle if the star is
west of the zenith.

Spherical triangle

The spherical triangle is more complicated than a plane triangle, but has particular defini-
tions as well. Please not that the sides of a spherical triangle are all on great circles. The
sum of all sides of the spherical triangle Ss is

Ss = a + b + c, (100)

with a, b, c < π. The spherical defect D is defined as

D = 2π − S. (101)

C Alexander Braun (2006-2010)


44 ENGO421: COORDINATE SYSTEMS
Figure 19: Left: Spherical triangle with the three interior angles (α, β, γ) and the three
sides angles (a, b, c). Right: Example scenario to derive the shortest distance between
Winnipeg and the Channel Islands.

Similarly, the sum of the interior angles Sa is defined as


Sa = α + β + γ, (102)
with π < Sa < 3π. The spherical access is defined as
E = Sa − π. (103)
The next definition is comparable to the planar sine theorem, which reads
sin α sin β sin γ
= = (104)
a b c
The planar sides become segments of great circles on the sphere. The following table
compares the planar with the spherical case.
planar spherical

angles: α + β + γ = 180 α + β + γ 6= 180◦
sides: a, b, c are distances a, b, c are spherical arcs measured in
measured in linear angular units
units
cosine formulae a2 = b2 + c2 − 2bc cos α cos a = cos b cos c + sin b sin c cos α
for the sides: cos b = cos c cos a + sin c sin a cos β
cos c = cos a cos b + sin a sin b cos γ
cosine formulae cos α = cos β cos γ + sin β sin γ cos a
for the interior angles: cos β = cos γ cos α + sin γ sin α cos b
cos γ = cos α cos β + sin α sin β cos c
sin β sin γ β γ
sine formula: sin α
a
= b = c sin α
sin a
= sin
sin b
= sin
sin c
sine-cosine formulae: sin a cos β = cos b sin c −
sin b cos c cos α
sin b cos γ = cos c sin a −
sin c cos a cos β
sin c cos α = cos a sin b −
sin a cos b cos γ

C Alexander Braun (2006-2010)


45 ENGO421: COORDINATE SYSTEMS
One of the most important applications of spherical trigonometry is the calculation of
distances on the sphere, e.g. to find the shortest distance between two places. Let us
consider an example, a pilot plans to fly from Winnipeg to the Channel Islands of the UK,
which are considered to be on the same latitude of 50o N . The longitude is 97o W and 2o W ,
respectively. Let us consider that the pilot knows that the shortest distance on a sphere is
the always along the great circle through the two points. Figure 19 outlines the scenario
and shows the angles used herein. Using the cosine law of spherical trigonometry:
cos c = cos a cos b + sin a sin b cos γcos2 40o + sin2 40o cos95o = 56.58o (105)
The distance in km along a great circle can be computed using the relation
2π RE /360o = 111km/1o (106)
Using the above relation, the distance between Winnipeg and the Channel Islands along
the great circle is 6280km. If the pilot would fly in easterly direction along the parallel,
the distance would be 95o ∗ 111km/1o ∗ cos 50o = 6778km. Comparing the two distances,
it is about 7.5% shorter to fly along the great circle. In order to derive the direction of the
great circle starting at Winnipeg, we can use the since rule to derive the heading of the
aircraft.
sin β sin γ
= (107)
sin b sin c
sin 40o sin 95o
sin β = = 0.77 (108)
sin 56.58o
Applying the arcsin operator to the last equation shows that two angles solve this problem,
50.1o or 129.9o . Common sense allows us to determine that the heading should be 50.1o . If
in doubt, we could also derive the angle b. The azimuth the aircraft start heading is then
360o − 50.1o = 309.9o . Note that this is about 40o north of the course along the parallel,
which would be 270o .

Astronomical triangle connecting H and HA

The coordinate pairs of the hour angle system HA (δ, h) and the horizon system H (A, a, z)
can be expressed using the concept of the astronomical triangle. This transformation is
very useful as it relates the measurements, which are taken in the horizon system H, to
the hour angle system, which can the nbe used to transform into the remaining celestial
systems E and RA. Figure 20 illustrates the relations between the parameters. Based on
the figure, we can obtain three equations for the unknowns (δ, h) based on the know
parameters (A, a, z, Φ).
cos δ sin h = − sin z sin A (109)
sin δ = cos z sin Φ + sin z cos A cos Φ (110)
cos δ cos h = cos z cos Φ − sin z cos A sin Φ (111)
The declination δ can be directly derived from the second equation. The hour angle h can
be derived by dividing the first equation by the third.
sin h − sin z sin A
= = tan h (112)
cos h cos z cos Φ − sin z cos A sin Φ

C Alexander Braun (2006-2010)


46 ENGO421: COORDINATE SYSTEMS
The inverse transformation has the know parameters (δ, h, Φ) and the unknowns (A, a, z).
The equations are as follows:
sin z sin A = cos δ sin h (113)
cos z = sin δ sin Φ + cos δ cos h cos Φ (114)
sin z cos A = sin δ cos Φ − cos δ cos h sin Φ (115)
The zenith angle z can be obtained directly from the second equation, and the azimuth
from the first and third.
sin h
tan A∗ = (116)
sin Φ + cos h − cos Φ tan δ
The azimuth is denoted by A∗ as there are four solutions, one in each quadrant. The
quadrant of the azimuth is determined by its sine and the signs of its cosine.

Time systems used in astronomic azimuth determination

The time scales used in astronomic azimuth determination are represented in the scheme
below. A more comprehensive description of the time systems and their relations is pre-
sented in chapter 4.

with:
αm = the right ascension of the fictitious Sun
= 12h 38m 45.836s + 8640184.542sT + 0.0929s T 2
Eq.E. = difference between the true and mean equinox.
T = in Julian centuries

Sidereal time and hour angle

Sidereal time and the hour angle are related through the right ascension angle, e.g. ST =
α + h. Graphically, this can also be expressed as in the following Figure 20.
From the figures, we find the following equation for the hour angle h of a celestial body:
LAST
z }| {
h
h = UT + (αm − 12 ) + Eq.E +Λ −α∗ .
| {z }
GAST

C Alexander Braun (2006-2010)


47 ENGO421: COORDINATE SYSTEMS
Spherical trigonometry -
hour angle and altitude method
North Pole
h
LAST

90
GA j

−Φ
ST
Aij
q *
S∗ z=90-a
-R
Rj β

i
Λ=0
δ A* Φ
Equin
ox

α Λ

Figure 20: Astronomic azimuth determination, parameters used in the hour angle method
and altitude method.

Determination of the astronomic azimuth

It is important to be able to determine the azimuth of objects on the Earth in many navi-
gation applications, e.g. the heading of the aircraft discussed in the previous section. Two
main methods for azimuth determinations are be considered: the hour angle method (of-
ten used as observations to Polaris) and the altitude method (often used as observations
to the Sun). A detailed description of both methods can be found in Thomson (1978). The
following sections are taken from a document prepared for the lab assignment - Astronom-
ical Azimuth Determination - by Drs. R. Grebnitcharsky and N. Sneeuw.

Definition: An astronomic azimuth is defined as the angle between the astronomic merid-
ian plane of a point i and the astronomic normal section through i and another point
j.

The required angles are shown in Figure 21.


To determine the astronomic azimuth of a line ij on the Earth we need:

• azimuth to a celestial body at time T ;


• measure the horizontal angle between the celestial body and the terrestrial reference
object.

The advantage of the hour angle method is that the observer has only to observe the star
as it coincides with the vertical wire of the telescope. No zenith distance (altitude) is
necessary. In this case the atmospheric refraction has no effect. The main disadvantage is
the need for a time registering device and a good knowledge of the longitude.
The advantage of the altitude method is that an accurate time device and precise longi-
tude are not necessary. The disadvantages of this method are that it is affected by the

C Alexander Braun (2006-2010)


48 ENGO421: COORDINATE SYSTEMS
Figure 21: Astronomic azimuth determination.

astronomic refraction and the fact that the star should coincide both with the horizontal
and the vertical wires. The altitude method is less accurate compared to the hour angle
method. They have the following precision: up to 1.5” for the hour angular method and
up to 5” for the altitude method.
All parameters and measurements, necessary for the hour angular method and altitude
method are represented graphically in the following figure.

3.9 Hour angle method


• From the sine formula applied to the co-latitude 90◦ − δ and the zenith distance z we
have:
sin (90◦ − δ) sin z cos δ sin h
= ⇒ sin β =
sin β sin h sin z
• From the cosine formula applied to 90◦ − δ we have:

cos (90◦ − δ) = cos z cos (90◦ − Φ) + sin z sin (90◦ − Φ) cos β


⇒ sin δ = cos z sin Φ + sin z cos Φ cos β
sin δ − cos z sin Φ
⇒ cos β =
sin z cos Φ

• From both sine and cosine formulas it is valid that:


cos δ sin h sin z cos Φ
tan β =
sin z (sin δ − cos z sin Φ)

• From another side the cosine formula for z gives:

cos z = cos (90◦ − Φ) cos (90◦ − δ) + sin (90◦ − Φ) sin (90◦ − δ) cos h
⇒ cos z = sin Φ sin δ + cos Φ cos δ cos h

C Alexander Braun (2006-2010)


49 ENGO421: COORDINATE SYSTEMS
• Substituting cos z in the equation for tan β and after some simplifications we have:
cos δ sin h cos Φ
tan β =
sin δ − (sin Φ sin δ + cos Φ cos δ cos h) sin Φ
cos δ sin h cos Φ
tan β = 2
sin δ(1 − sin Φ) − cos Φ cos δ cos h sin Φ
cos δ sin h cos Φ
tan β =
cos Φ(cos Φ sin δ − cos δ cos h) sin Φ
sin h
or tan β =
tan δ cos Φ − sin Φ cos h
• Finally, taking into account that tan β = tan (360◦ − A∗ ) = tan −A∗ = − tan A∗ , we
get for the azimuth of the star:
sin h
tan A∗ =
sin Φ cos h − tan δ cos Φ
The corresponding differential formula of tan A∗ displays the effect of errors in Φ and h
(see the derivation with the altitude method)on the azimuth:
A∗ = sin A∗ cot zΦ + cos Φ(tan Φ − cos A∗ cot z)h
sin A∗ cos δ cos q
= Φ+ h
tan z sin z
. For instance

• when A∗ = 0◦ or A∗ = 180◦ the effect of Φ is eliminated. This is the elongation case.


• when tan Φ = cos A∗ cot z the effect of h is eliminated.

A special case occurs for circumpolar stars, like Polaris (α Ursae Minoris). For Φ > 15 ◦ it
is easily visible and not affected by the atmospheric refraction. Polaris has the following
advantages:

• A∗ ≈ 0◦ the error Φ is minimized;


• For Polaris we have δ ≈ 90◦ and the effect of the hour angle error will be eliminated.

These two advantages of Polaris make this star the best choice for the hour angle method.

The error budget for the hour angle method using Polaris:

• pointing errors
• timing errors in local time device → h
• longitude errors → h
• latitude errors → Φ
• errors in interpolation of the right ascension α∗ and the declination δ of Polaris. They
are tabulated in different star catalogs and almanacs.

C Alexander Braun (2006-2010)


50 ENGO421: COORDINATE SYSTEMS
Observation procedure

The suggested observation procedure is according to Mueller (1977):

1. Direct to Earth object, record horizontal circle reading (HCR).


2. Direct to Polaris, record HCR and time T .
3. Repeat step 2).
4. Reverse the telescope and repeat steps 1–3).

Repeat the entire procedure eight times, having eight sets of observations.
To get a more precise azimuth a striding level can be used. The correction of the azimuth
because of the levelling is:

d00
00
∆A = ((w + w 0 ) − (e + e0 )) cot z,
4
in which

d00 is the value of each division of the striding level in arc seconds.
e, w are the readings of both ends of striding level in direct observations.
e , w0
0
are the readings of both ends of striding level in reverse observations.

Computation procedure

The computation procedure is described in Mueller (1977) and Thomson (1978).

1. compute the hour angle for every mean direct and reverse readings.
2. compute the astronomic azimuth corresponding to every mean direct and reverse
time readings.
3. compute the astronomic azimuth of the terrestrial line from mean direct and mean
reverse HCR to Polaris and the mean readings to the Earth object.
4. compute the statistics for all eight set of observations. Every one consists of two
(direct and reverse) azimuths. Mean value, standard deviation of one observation
and standard deviation of the mean value are the necessary statistics.

The astronomic azimuth by the hour angle method is used for all types of astronomic work
because of the high accuracy that can be reached.

C Alexander Braun (2006-2010)


51 ENGO421: COORDINATE SYSTEMS
3.10 Astronomic azimuth determination by the altitude method
Applying the spherical cosine formula for 90◦ − δ and z = 90◦ − a we will have:
cos (90◦ − δ) = sin (90◦ − Φ) sin (90◦ − a) cos β + cos (90◦ − Φ) cos (90◦ − a).
Taking into account that cos β = cos (360◦ − A∗ ) = cos A∗ we have:
sin δ = cos Φ cos a cos A∗ + sin Φ sin a.
Finally, for the astronomical azimuth of the celestial body we get:
sin δ − sin Φ sin a
cos A∗ =
cos a cos Φ

with a = 90 − z is the altitude of the celestial body.
The measured values are the zenith distance z or the altitude angle a. The values for the
declination δ of the celestial body are obtained from catalogs, ephemeris or almanacs. The
time is necessary with an accuracy to the nearest minute to determine δ.
Differentiating the equation for cos A∗ leads to:
sin A∗ A∗ = (tan a − cos A∗ tan Φ)Φ + (tan Φ − cos A∗ tan a)a.

Now we are able to investigate the effect of different errors.


• Let us consider one sine-cosine formula for the spherical triangle in Figure 20.
sin (90◦ − δ) cos h = cos (90◦ − a) sin (90◦ − Φ) − sin (90◦ − a) cos (90◦ − Φ) cos A∗ .
If h = 90◦ = 6h or h = 270◦ = 18h we will have:
sin a cos Φ = cos a sin Φ cos A∗ =⇒ tan a − cos A∗ tan Φ = 0.
This fact means that the effect of Φ will be eliminated.
• Let us consider another sine-cosine formula for the spherical triangle in Figure 20.
sin (90◦ − δ) cos q = cos (90◦ − Φ) sin (90◦ − a) − sin (90◦ − Φ) cos (90◦ − a) cos A∗ .
If the parallactic angle of the celestial body q = 90◦ = 6h or q = 270◦ = 18h we will
have:
sin Φ cos a = cos Φ sin a cos A∗ =⇒ tan Φ − cos A∗ tan a = 0.
This fact means that the effect of altitude errors is eliminated. It happens when the
celestial body is at elongation.
• Both errors — Φ and a — can not be simultaneously eliminated. But their effect
could be minimized if two stars are observed. The two stars should have a1 = a2 and
A1 = 360◦ − A2 .
For azimuth determination by the altitude method the observations of the Sun are used
very often. Using a 100 theodolite the standard deviation of the determined astronomic az-
imuth could be 2000 . With special solar attachments (solar prism) the accuracy of determine
azimuth could be 500 .
The suggested observation procedure consists of two morning and two afternoon deter-
minations. The best choice for the latitude a is 30◦ < a < 40◦ , but never a < 20◦ and
a > 50◦ .

C Alexander Braun (2006-2010)


52 ENGO421: COORDINATE SYSTEMS
Observation procedure for azimuth determination by altitude observations of the Sun

The following observation procedure has been proposed in Mueller (1977) and Thomson
(1978):

1. Direct to terrestrial object, record horizontal circle reading (HCR) and level (plate or
striding level) readings.
2. Direct to the Sun, record the HCR and the vertical circle reading (VCR). Record the
time to the nearest minute.
3. Reverse the telescope. Direct to the Sun, record HCR and VCR and time to the nearest
minute. Record the level readings.
4. Reverse on the Earth object, record HCR and level readings.
5. Record the temperature and pressure.

Computation procedure for azimuth determination by altitude observations of the


Sun

The computation procedure could be found in Mueller (1977) and Thomson (1978).

1. Compute the mean direct and reverse HCRs to the Earth object and to the Sun
2. Correct the horizontal directions in 1)) for the readings of the striding level. Use the
formula:
d00
∆A00 = ((w + w 0 ) − (e + e0 )) cot z,
4

which is the same as in the hour angle method.


3. Compute the angle between the Sun and Earth object.
4. Compute the direct and reverse zenith distance (altitude) to the Sun. Correct the
mean values for the effect of astronomical refraction.
5. Determine the time of observation to be able to obtain the tabulated declination δ
values of the Sun.
6. Compute the azimuth of the Sun, using the formula for cos Asun .
7. Using the angle from 3) and the azimuth of the Sun from 6) to compute the azimuth
to the Earth object.

The altitude method for astronomic azimuth determination by observations to the Sun has
an accuracy, which will be enough for purposes requiring 500 .

C Alexander Braun (2006-2010)


53 ENGO421: COORDINATE SYSTEMS
References

Thomson,D.B.(1978). ”Introduction to Geodetic Astronomy”,Lecture Notes


No.49,Department of Surveying Engineering,University of New Brunswick, 175pp.
Mueller,I.I.(1977). ”Spherical and Practical Astronomy as Applied to Geodesy”,Ungar,New
York,615pp.

C Alexander Braun (2006-2010)


54 ENGO421: COORDINATE SYSTEMS
4 Time Systems
In the previous chapters, the celestial coordinate systems involved the parameter time,
e.g. in the hour angle or right ascension angle definition. Earth rotation can only be
included if time is considered properly. This chapter introduces a more detailed description
of time. Time is the fourth dimension in geomatics engineering, geodesy and our daily
life. Everybody knows time and everybody has a feeling of what it means. In a way,
time can be considered a natural coordinate, e.g. comparable to the gravity vector, which
tells us where up and down is, time tells us the sequence of events. There would be
no navigation or geodesy possible without knowing time. Time is very likely the oldest
physical parameter mankind understood and tried to develop, as it is a fundamental part
of human communication and information exchange. Time can be understood in three
different ways,

1. as an epoch, or an exact point in time,


2. as a time interval or period,
3. and as a time scale.

Examples for the three definitions are the following three questions:

1. What time is it? Answer: Time epoch 11 : 25 AM.


2. How much time is left before the break? Answer: Time interval: 20 min.
3. How old are you? Time scale: 20 years or 6.31152 × 108 sec.

Time is also employed in several geodetic applications ranging from the definition of 1
metre, over the travel-time of a GPS signal to derive a distance, to the transformation of
inertial and terrestrial coordinate systems. For instance, the distance an electromagnetic
wave travels can be expressed by
d = c t → ∆d = c ∆t (117)
Using this equation to estimate the required time precision for a distance precision of 1cm
results in 33ps = 3.3×10−11 s. As satellites travel with a ground speed of 7−8 km/s, precise
timing is also mandatory to monitor their position and predict their orbital path. Time
also transforms inertial systems (e.g. the ecliptic system) into terrestrial systems (e.g. the
horizon system). While the later is fixed with the Earth and cannot observe Earth rotation
as it rotates with the Earth, the ecliptic system is considered the most stable inertial system
and does allow the observation of Earth rotation. The transformation parameter of major
importance is the time or the velocity of Earth rotation.
As a consequence of the above considerations, three different time systems will be intro-
duced.

1. Sidereal time - refers to the vernal equinox, or stars


2. Solar/universal time - refers to the Sun
3. Atomic time - refers to the oscillation period of atoms

C Alexander Braun (2006-2010)


55 ENGO421: COORDINATE SYSTEMS
While the first two systems can be considered natural time systems as they refer to natural
phenomena which are meaningful to us, the third one is more stable but is difficult to
understand for us. All three time systems will be discussed in the following and also
their relation and the required transformations will be introduced. Talking about time
also implies calendar time and time zones, which are arbitrary definitions using the time
systems above. These will be introduced as well in combination with Julian Date and U T C
time transmitters.
Before the time systems will be discussed, it is important to understand what characteris-
tics a time system should have. Certainly, it is of advantage to develop a time system which
makes sense to mankind everywhere on Earth. It should be meaningful and it should not
diverge from what we think is useful, e.g. that noon is when the Sun reaches the highest
elevation above the observers location. Keeping it meaningful, however, implies that we
must accept changes in our time system as the natural processes which define the time
(vernal equinox location, Sun location) have variations. It implies that one second today
is not equal to a second next year. This characteristic is certainly not desired as it means
that the time system is not stable, or in other words, the time scale changes. Unfortunately,
both natural time systems cannot be considered stable as the vernal equinox and the Sun
are not fixed relative to the Earth and variations of the Earth rotation rate further compli-
cate this issue. On the other hand, a perfectly stable time systems based on atomic time
has no meaning to us, as we cannot not comprehend what roughly 9 billion oscillation
of an atom mean. Later we will see that atomic time can serve as the time scale and the
natural time systems give us the meaningfulness. A time system must also be accessible,
it does not make sense to carry a T-2 theodolite with us and do star/Sun observation to
obtain the time. In summary, a time system must have the following characteristics:

• Stability: 1 seconds today is as long as 1 second tomorrow.


• Accessibility: Time can be obtained anytime and everywhere.
• Long-term availability: Time system should work during human civilization of the
Earth or planets.
• Meaningfulness: Time system needs to make sense, noon sees daylight, midnight
remains at night.

4.1 Sidereal Time


Sidereal time is defined as the angle between the given meridian and the vernal equinox
(), all in the equatorial plane. The angle is expressed in units of time, here, sidereal hours
or sidereal days. One sidereal day is defined as the time period between two consecutive
transits of the through the meridian of choice. In principle, any star could be used as
well, but the intrinsic motion of the star is not known or irregular, so that the becomes
the preferred reference. Assuming that the position of the is fixed, one revolution of the
Earth corresponds to one sidereal day. The assumption that the is fixed is not always true,
so that we defined the actual or true position of the equinox T and the mean position of
the equinox M . Depending on the choice of the , sidereal time ST becomes the Actual
Sidereal Time (AST) or the Mean Sidereal Time (MST). This holds for every meridian, but
if a particular meridian is selected, e.g. the local meridian, ST becomes Local Sidereal Time
(LST) or, in case of the Greenwich meridian, ST becomes Greenwich Sidereal Time (GST).

C Alexander Braun (2006-2010)


56 ENGO421: COORDINATE SYSTEMS
local meridian Greenwich meridian
true equinox T LAST GAST
mean equinox M LMST GMST

Table 1: Sidereal time systems.

Combined, this results in four different sidereal time systems as shown in the following
table.
A graphical interpretation of ST can be achieved from looking at the astronomical triangle
discussed in chapter 3, and shown in Figure 20. Projected onto a plane with the NCP in
the centre, Figure 22 shows the relation between ST , α, h, and λ. In the previous chapter

Figure 22: Sidereal time systems, right ascension angle and hour angle.

3, ST was defined as the sum of right ascension angle and hour angle, from Figure 22 we
can now identify the different ST systems in relation to α and h.

LAST = α + h (118)
GAST = LAST − Λ = α + h − Λ (119)

Based on the above equation, the astronomical longitude Λ can be derived from measuring
the hour angle h to the star, taking the right ascension angle α from a catalog, and having
the time GAST . The hour angle becomes zero, when the local meridian is passed by the
star, this is called upper culmination. As we identified earlier, the moves due to nutation

C Alexander Braun (2006-2010)


57 ENGO421: COORDINATE SYSTEMS
along the ecliptic plane. This results in a time dependent value for GAST and LAST . The
angle change in longitude along the ecliptic plane which results from the motion of the
equinox is ∆ψ. As ST is measured in the the celestial equator plane, this angle ∆ψ needs to
be projected onto the celestial equator by the angle of the ecliptic , so that the difference
between actual/true and mean ST can be expressed by the Equation of the Equinox
Eqn. Equinox = ∆ψ cos  = LAST − LM ST = GAST − GM ST (120)
The magnitude of the Eqn. of Equinox is about 1s, however, this can be important for
high precision time keeping. It is thus corrected whenever possible. Figure 23 shows the
variations of the Eqn. of Equinox after Müller. After correcting ST by the Eqn. of Equinox,

Figure 23: Equation of Equinox variations after Müller.

ST is reasonable stable, however, not perfect. A more meaningful time will be introduced
now as solar or universal time.

4.2 Solar or Universal Time


Instead of using the as a reference, solar or universal time takes the Sun as the reference,
Hence, two consecutive transits of the Sun through the local meridian define one solar day.
For convenience and meaningfulness, 12 hours are added to the transit epoch in order to
keep noon or 12:00 at the time of the upper culmination of the Sun with respect to the local
meridian. The definition of solar time is quite simple, but needs significant corrections as
the Sun-Earth/Moon system moves relative to each other and has temporal variations
along the orbit. The major effect results from the fact that the Earth/Moon system orbits
the Sun on an elliptical orbit as outlined in Figure 24. Two different types of motion must

C Alexander Braun (2006-2010)


58 ENGO421: COORDINATE SYSTEMS
Figure 24: Sun-Earth/Moon orbital motion and Kepler’s 2nd law.

be considered, the first one is the Earth revolution about its spin axis, the second is the the
Earth orbital motion about the Sun. While the first motion covers one sidereal day per full
revolution, there will be an additional revolution angle in order to have the Sun transiting
the local meridian. This is due to the fact that the Sun moved relative to the Earth as it is
not fixed like the stars. The orbital motion of the Earth about the Sun introduces one extra
day per year, which accounts for this extra motion. Consequently, this extra day represents
the difference between sidereal and solar time. A solar day is longer than a sidereal day as
the following equation demonstrates:
360o
1 mean solar day = = 0.98565 o /day → 3min56.33sec per day (121)
365.242 days
1 mean solar day = 1 mean sidereal day + 3min56.33sec (122)
366.242
F = = 1.00273791 scale factor (123)
365.242
This scale factor was previously used to interpolate the R factor in the transformation from
U T 1 to GAST . As in the previous chapter on sidereal time, solar time can be referred to
the local or Greenwich meridian, thus, local solar time becomes

LT = hSun + 12h. (124)

To ensure meaningfulness, 12 hours are added, otherwise LT = 0 : 00 or midnight when


the Sun is at the upper culmination. Using the Greenwich meridian as the reference merid-
ian, LT becomes U T .
U T = hGM
Sun + 12h = LT − Λ (125)
Figure 25 shows these relations. As indicated before, the true position of the Sun moves
significantly due to the relative motion to the Earth. This results in an unstable time system
as outlined in Figure 24. The Earth moves faster along the orbital path about the Sun if
it is closest to the Sun (Perihelion), it moves slower if it is farthest away (Aphelion). This
is also expressed by Kepler’s 2nd law, which states that the line joining the Earth and the
Sun sweeps out equal areas during equal intervals of time. This is also known as the law
of equal areas. Due to this inhomogeneous motion, the apparent path of the Sun leads
to a time lag which is different every day depending on the constellation between Sun

C Alexander Braun (2006-2010)


59 ENGO421: COORDINATE SYSTEMS
Figure 25: Solar time systems, right ascension angle, and hour angle.

and Earth. The average is, however, 3min56.33sec. Let us introduce a fictions Sun, which
would lead to a time lag of exactly 3min56.33sec every day. This Sun is also called mean
Sun and has a homogeneous speed relative to the Earth. As we defined the Eqn. of Equinox
for the difference between true and mean equinox, we defined now the Equation of Time
as the difference between the true and the mean Sun.

Eqn. of Time = αSun


¯ − αSun = hSun − hSun
¯ (126)

The magnitude of the Eqn. of Time is about ±15min, compared to 1s of the Eqn. of
Equinox. The demonstrates that sidereal time is more stable than U T . Once U T is cor-
rected for the heterogeneous motion of the Sun, a more stable solar time is obtained, called
U T 0.
U T 0 = U T − Eqn. of Time = hGM¯ + 12h
Sun (127)
Now, universal time U T 0 is reasonably stable, but there is another correction which needs
to be applied. This correction accounts for polar motion, which changes the orientation
of the Earth’ spin axis and hence, the orientation of the Greenwich meridian. A mean
Greenwich meridian is introduced and Greenwich was selected due to the long-term ob-
servations of the Greenwich Observatory. It was a logical choice to set the zero meridian
to Greenwich. The spin axis of the Earth changes orientation by about 0.3i arcsec per year.
The new solar time system is called U T 1 which is identical with GM T , Greenwich Mean
Time.
¯
U T 1 = GM T = U T 0 + ∆Λp = U T 0 − (xp sin Λ + yp cos Λ) tan Φ = hGM
¯ + 12h
Sun (128)

C Alexander Braun (2006-2010)


60 ENGO421: COORDINATE SYSTEMS
The polar motion parameters xp and yp are obtained from observations. The difference
between U T 0 and U T 1 is shown in Figure 26. After this correction for polar motion,

Figure 26: Relation between UT1 and UT0.

U T 1 is quite stable, however, there are periodic and episodic variations of the spin axis
direction due to mass transfer processes on the Earth surface and in the Earth’ interior.
As these variations are currently unpredictable, and mostly cause very small effects, the
correction for such variations is not applied. If applied, the new solar time is called U T 2.
Examples for processes changing the spin axis orientation and the rotation rate, and thus
the length of a solar day, include earthquakes, mantle convection, floods, and deglaciation
of major ice sheets. For instance, the Sumatra-Andaman Island earthquake decreased the
Earth’ rotation rate by 2.5 µs, and displaced the spin axis by 2.5cm. Consequently, a solar
day is not constant and the length of day varies by milliseconds.

4.3 Conversion between Sidereal and Solar Time


It is useful to develop relations between the two natural time systems, before we introduce
an ultra-stable time system, atomic time. One tropical year is defined as the time inter-
val between two consecutive transits of the mean Sun through the mean vernal equinox,
or 365.242198797 mean solar days. In order to convert from sidereal to solar time, the
difference between the mean equinox and the mean Sun is required. From the previous
sections, we obtain:
GM ¯
GM ST = αSun
¯ + hSun
¯ = αSun
¯ + U T 1 − 12h (129)

The right ascension angle of the mean/fictions Sun can be obtained from Sun catalogs or it
can be estimated using the following convention of the International Astronomical Union
of 1976:
h m s s 2
αSun
¯ = 18 41 50.54841 + 8640184.812866 T + 0.0929 (T ), (130)

C Alexander Braun (2006-2010)


61 ENGO421: COORDINATE SYSTEMS
with T = time since January 1, 12:00 noon, 2000.
d
T = , (131)
36525
with d = JD − 2451545.0, JD = Julian Date. GM ST can then be expressed by

GM ST = U T 1 + 24110.54841 + 8640184.812866 T + 0.0929 s (T 2 ) (132)

or
GAST = U T 1 + (αSun
¯ − 12h + Eqn. of Equinox) = U T 1 + R (133)
The parameter R can be derived from star catalogs, where it is available in intervals of 6
or 12 hours. In case that R is required at times other than available from the catalog, R
needs to be interpolated. If the processes involved would be linear, this would be simple,
but they are not. Over short time periods, however, we can assume that the change of R
is linear. Assuming that R is needed at U T 1 to derive GAST = U T 1 + R. Then, R0 at
U T 10 (closest epoch to U T 1) is obtained from the catalog and GAST can be derived from

GAST = R0 + U T 10 + (U T 1 − U T 10 ) F (134)
U T 1 = U T 10 + (GAST − R0 − U T 10 )/F. (135)

4.4 Time Zones and Calendar Time


Time Zones

Solar time and sidereal time are longitude dependent, it is obvious that Earth rotation
changes the time and GM T can only remain meaningful at the Greenwich meridian. As the
Earth spins by 15o every hour, for every time zone of 15o , a mean meridian is defined and
an integer number of hours is added or subtracted from GM T or U T 1 to obtain meaningful
time at the particular time zone. In order to get U T 1 in the time zone XY Z, the zonal
correction ∆Z = −11, −10, −9...0, +1, +2... + 11 is applied, so that U T 1 = XY Z + ∆Z.
The local sidereal time in Alberta (Mountain Standard Time, MST) can then be derived
from
LST = (M ST + ∆Z)/F + R0 + Λ (136)

Calendar Time

Calendar time was always important throughout human history as it is easier to remember
a month rather than remembering the day of year of a particular event. Julius Caesar
(46 B.C.) defined one year has 365 days. This is not accurate enough as we know that a
tropical year is longer than that. The Julian Century accounts for that, and 1JC = 36525 d.
Compared to a tropical year, which is 365.2422 d, there is a slight difference which results in
the fact that after 131 years, the Julian Calendar is out of sync by one day. It took another
1500 years before Pope Gregory corrected this to fix the date of Easter. He decided that
10 days in October 1582 must be skipped. Further, he changed the leap year rule from the
Julian Calendar to the Gregorian Calendar. The Julian Calendar added one extra day in
every 4th year, the Gregorian Calendar now adds one extra day every 4th year, but not if

C Alexander Braun (2006-2010)


62 ENGO421: COORDINATE SYSTEMS
the year is evenly divisible by 100, except if it is also divisible by 400. Hence, the years
2000 and 2400 are leap years, but 1900 and 2100 are not.
Civilian time counts the years, months and days based on the Gregorian Calendar. How-
ever, the length of one month is variable, e.g. February has 28 days and August 31 days.
For numerical reasons, this is difficult to implement in computer code. So, a time system
has been introduced, which continuously becomes larger and does use the time scale of
Julian days. The Julian Date started at noon January 1, -4712 (or 4713 B.C.). Every new
Julian day starts at noon. Conversion of Gregorian calendar date to Julian date for the
years 1801-2099 can be carried out with the following formula:

JD = 367K− < (7(K+ < (M + 9)/12 >))/4 > + < (275M )/9 > +I + 1721013.5 (137)
+U T 1/24 − 0.5sign(100K + M − 190002.5) + 0.5 (138)

where K is the year (1801 <= K <= 2099), M is the month (1 <= M <= 12), I is the
day of the month (1 <= I <= 31), and U T 1 is the universal time in hours (“<=” means
“less than or equal to”). The last two terms in the formula add up to zero for all dates after
1900 February 28, so these two terms can be omitted for subsequent dates. This formula
makes use of the sign and truncation functions described below:
The sign function serves to extract the algebraic sign from a number.
Example: sign(247) = 1; sign(−6.28) = −1.
The truncation function < > extracts the integral part of a number.
Example: < 17.835 >= 17; < −3.14 >= −3.
Example: Compute the JD corresponding to 1877 August 11, 7h30m UT. Substituting
K = 1877, M = 8, I = 11 and U T 1 = 7.5,

JD = 688859 − 3286 + 244 + 11 + 1721013.5 + 0.3125 + 0.5 + 0.5 = 2406842.8125 (139)

As the Julian Date for last 150 years always starts with 24...., the Modified Julian Data
(MJD) has been introduced to save some digits. MJD starts at Midnight 0:00 November
16-17, 1858, or
M JD = JD − 2400000.5 (140)
As an example, the Gregorian Calendar time of 9:30, October 31, 2006, translates to
M JD = 54039.395833.

4.5 Atomic Time


The natural time systems have not been stable due to the natural processes and their
variations which served as the reference. It is therefore desirable to find a time scale
which is ultra-stable. A combination of the two time systems would then provide the
best available time system which fulfills all time keeping requirements listed before. The
historic definition of 1 second was the time 1/86400 of a mean solar day. As mentioned
before, a solar day is not constant as the Earth slowly decreases its spin rate. This led
to a redefinition of the duration of 1 second in 1967 by the Bureau International des
poid and measures. “One second is the duration of 9192631770 periods of the radiation
corresponding to the transition between the two hyperfine levels of the ground state of
1
33 Cs atom”. It is important to notice that this definition requires that the clock stands
still, otherwise relativistic effects must be accounted for. These effects are accounted for

C Alexander Braun (2006-2010)


63 ENGO421: COORDINATE SYSTEMS
in the atomic clocks on-board the GPS satellites. The new time system is called T AI
(Temps Atomique International). T AI is the weighted average of 200 atomic clocks in 30
different countries. GPS time is using atomic time as well, each satellite has an atomic
clock on-board. However, TGP S uses a composite clock with different settings and does
not represent T AI. There are offsets and variations on the µs level, which are included in
the GPS broadcast message.
A closer look on T AI and U T 1 reveals that the two times will diverge with time, because
the length of day changes, because the spin rate decreases, and because of variations in
the spin rate. In order to combine ultra-stable T AI with meaningful U T 1, the Universal
Time Coordinated (U T C) has been introduced. T AI gives the time scale for U T C, but it
is desired to keep U T C close to U T 1. This can be achieved by introducing leap seconds
according to the following rule, if

∆U T = U T C − U T 1 >= 0.9s (141)

Leap seconds are released by the International Earth Rotation Service (IERS) on Jan 1

Figure 27: Leap seconds as a function of T AI, U T C, and U T 1.

and Jul 1 every year. The leap second rule started in 1972 and n = 10 leap seconds were
introduced. Figure 27 shows the continuous change of U T 1 and the discrete change of
U T C which follows U T 1 through the application of leap seconds. On Jan 6, 1980, GPS
time was introduced and TGP S − T AI = 19s. Since then, TGP S remains parallel with T AI
and no leap seconds are applied to TGP S . This leads to a drift of TGP S with respect to
U T C and U T 1. The drift is included in the broadcast message of GPS satellites and the
coefficients are obtained from the following equation.

TGP S − U T C = n − f (A0 , A1 ), with f (A0 , A1 ) < µs (142)

In conclusion, U T C combines all required characteristics of time keeping, it is stable, it is


accessible, and it is meaningful. The next section will now show how to obtain U T C for
your daily work.

C Alexander Braun (2006-2010)


64 ENGO421: COORDINATE SYSTEMS
Figure 28: Time transmitter antenna array in Fort Collins, CO.

Time Transmitters

U T C needs to be made available at any time and anywhere on Earth. A number of time
transmitting antennas have been established to broadcast a radio signal which contains a
binary code of U T C. There are about one or two such antennas on every continent. The
North American time transmitter is located in Fort Collins, CO and transmits U T C on a
60Hz signal (Figure 28). The binary code is based on one bit every second, which means
1, if the signal is longer than 110ms, and 0 otherwise. The code required to read the binary
signal is in WWVB time code format as shown in Figure 29. This code does further allow
the receiver to obtain U T 1 as the correction to U T C is part of the signal. This signal is
received and decoded by radio-controlled watches or clocks. As the code and the signal
frequency change from station to station, most of these watches do not work on other
continents, unless a multi-time code is implemented. This concluded the chapter on time
systems, which provides the 4th dimension of coordinate systems.

C Alexander Braun (2006-2010)


65 ENGO421: COORDINATE SYSTEMS
Figure 29: Time transmitter WWVB time code format.

5 Terrestrial Coordinate Systems


In chapter 2, natural coordinates have been introduced and the coordinate frames LA and
CT have been established. Chapter 3 developed celestial coordinates which were also
based on natural coordinates such as the spin axis of the Earth, the vernal equinox or the
gravity vector. In this chapter, a new type of coordinate systems is introduced which even-
tually becomes independent from the natural coordinates and introduces a mathematical
reference surface called the ellipsoid which replaces the geoid in the previously discussed
astronomical coordinate frames LA and CT . These new ones are called geodetic coordi-
nate frames.
The question arises why do we make the transition from the natural/physical to mathe-
matical/geometrical coordinates? The function W (x, y, z) describing the gravity potential
has a complicated form. Hence, the geoid, given by the equation W0 = const., is a rather
complicated surface and can be described rigorously only by an infinite functional series.
Although modern space gravimetric satellite missions have led to a much more precise
knowledge of the geoid and the Earth gravity field, there are still uncertainties of several
centimeters, particularly in the polar regions. Similarly, the orientation changes of the
local astronomic frame cannot be expressed by simple analytical functions as they directly
depend on the orientation of the gravity vector and thus the density distribution within
the Earth. This means that it is difficult to connect measurements between two different
points. Therefore, the geoid is approximated by a surface with a simple mathematical

C Alexander Braun (2006-2010)


66 ENGO421: COORDINATE SYSTEMS
representation. This allows for computations which account for the differences between
this model surface and the actual geoid which can be added as corrections to the results
obtained from computations on the approximating surface. It also connects the orientation
of the local frame at one point to that at another by simple formulas. In this chapter, the
approximation of the geoid by a best fitting ellipsoid and the approximation of the local
astronomic frame by the corresponding local geodetic frame will be defined. The location
and orientation of this ellipsoid in space, the so-called datum problem, will be treated in
the next chapter.

Figure 30: Sphere, geoid, geoid undulations N and the best fitting ellipsoid (not to scale).

5.1 The Best Fitting Ellipsoid


The geoid is a irregular surface often compared to a potato, but it reflects the flattening
of the Earth. The simplest surface which can be considered as a geoid approximation is
the sphere. Locally it fits the geoid well and thus it can be used in certain cases as a
local approximation to the geoid. Globally, however, the sphere is a poor approximation
of the geoid. The difference between the geoid and the best fitting sphere will be as
much as 10km because the Earth’s flattening which causes the equatorial and polar radius
difference of about 21km. A better approximation is provided by a rotational or biaxial
ellipsoid which is depicted in Figure 30. It is a good approximation of the geoid both
locally and globally because the distance N between the geoid and the best fitting ellipsoid
is only about ±100m, or more precisely

−110m < N < 90m. (143)

The physical explanation of this phenomena was given by Newton (1642-1727) and Huy-
gens (1629-1695) who independently developed the theory of equilibrium figures and
showed that a rotational ellipsoid is obtained as the equilibrium figure of a homogeneous,

C Alexander Braun (2006-2010)


67 ENGO421: COORDINATE SYSTEMS
rotating Earth. A rotational ellipsoid is thus a good approximation of the geoid from both
a theoretical and a practical point of view; its geometry is relatively simple, and so are
the computations on the ellipsoidal surface, although slightly more complicated than on
the sphere. In order to determine the best fitting ellipsoid that can be used as a reference
surface for computations, two approaches are possible. The first one is purely geometri-
cal and minimizes the orthogonal distance between geoid and ellipsoid globally. Such an
ellipsoid is called the Geodetic Reference Ellipsoid (GRE) in the following. To define the
GRE, only two ellipsoidal parameters are necessary, e.g. the two semi-axes a and b, or one
of the semi-axis and the flattening parameter f of the ellipsoid. Although such an ellipsoid
is very useful for geodetic computations with distances and angles measured on the sur-
face, it cannot be used for measurements that are affected by the gravity field of the Earth.
In this case, the gravity potential on the ellipsoidal surface and external is required. This
so called normal potential U can be defined by fixing the parameters for the mass of the
Earth (ME ) and its rotation rate (ωE ). The resulting gravity on the surface of the ellipsoid
is called normal gravity and is denoted by ~γ . The reference defined by a, b, ME , and ωE
is called a Geodetic Reference System (GRS) because it defines a geometrical as well as a
physical reference for geodetic measurements. The current systems that are widely used
are the GRS1980 and WGS1984. The best fitting ellipsoid can be defined geometrically as
an ellipsoid satisfying the condition
X Z
Ni2 = M in =⇒ N 2 dS = M in, (144)
i=1 S

with geoid undulations N and surface S. It thus minimizes the deviations between the
geoid and the ellipsoid in a global sense. In practice, a best fitting ellipsoid can be com-
puted using least squares methods. The distance N between the geoid and the best fitting
ellipsoid can be computed using the following fundamental formula:
N =h−H (145)
or
Geoidal undulation = ellipsoidal height - orthometric height

Figure 31: Geoid undulations N , ellipsoidal height h, orthometric height H, and vertical
deflections θ.

Figure 31 shows that the equations N = h − H is an approximation only, i.e. the angle θ
has been considered as zero. This angle results from the fact that the gravity vector normal

C Alexander Braun (2006-2010)


68 ENGO421: COORDINATE SYSTEMS
on the geoid is not parallel to the normal gravity vector which is normal to the ellipsoid.
The difference in directions is described by the deflection of the vertical θ. However, the
resulting errors are negligible for all practical purposes because θ is rarely larger than one
arc-minute. Assuming θ = 10 and h = 2000m, the difference between h and (H + N ) is less
than 0.1mm.
Historically, best fitting ellipsoids were frequently determined for continents or parts of
them, i.e. they were ellipsoids best fitting to only parts of the Earth’s figure. These were
called locally best fitting ellipsoids, because measurements were only available in a re-
stricted area and the fit was therefore not optimal in a global sense (Figure 32).

Figure 32: Locally best fitting ellipsoid.

Consequently, different continents used different locally best fitting ellipsoids which min-
imize the sum of N 2 only over the particular continent. Examples of locally best fitting
ellipsoids are:

• Bessel (1841) for Europe


• Clarke (1866) and Hayford (1909) for North America
• Krassowsky (1940) for the former USSR

Examples of globally best fitting ellipsoids are:

• GRS of 1967 and 1980


• WGS of 1960, 1966, 1972, and 1984

The parameters of the most commonly used reference ellipsoids are given in Table 2.
While the formula N = h − H and Figure 31 suggest that the determination of geoid
undulations N is a purely geometrical problem, this is only true for ellipsoidal heights h
which are determined as distances along the normal to the ellipsoid. It is not true for
orthometric heights H which are a function of the gravity potential W (x, y, z) and the
direction of the plumbline or the gravity vector. The geometry problem is therefore to
determine h, e.g. from GPS, the physical problem is to determine H using levelling and
gravimetry. Both are related through N .

C Alexander Braun (2006-2010)


69 ENGO421: COORDINATE SYSTEMS
Airy 6377563.396 299.3249646 573.604 0.11960023
Australian National 6378160.0 298.25 -23.0 -0.00081204
Bessel 1841 6377397.155 299.1528128 739.845 0.10037483
Bessel 1841 (Nambia) 6377483.865 299.1528128 653.135 0.10037483
Clarke 1866 6378206.4 294.9786982 -69.4 -0.37264639
Clarke 1880 6378249.145 293.465 -112.145 -0.54750714
Everest 6377276.345 300.8017 860.655 0.28361368
Fischer 1960 (Mercury) 6378166.0 298.3 -29.0 0.00480795
Fischer 1968 6378150.0 298.3 -13.0 0.00480795
GRS-1967 6378160.0 298.247167427 -23.0 -0.00113048
GRS-1980 6378137 298.257222101 0.0 -0.00000016
Helmert 1906 6378200.0 298.3 -63.0 0.00480795
Hough 6378270.0 297.0 -133.0 -0.14192702
International 6378388.0 297.0 -251.0.0 -0.14192702
Krassovsky 6378245.0 298.3 -108.0 0.00480795
Modified Airy 6377340.189 299.3249646 796.811 0.11960023
Modified Everest 6377304.063 300.8017 832.937 0.28361368
Modified Fischer 1960 6378155.0 298.3 -18.0 0.00480795
South American 1969 6378160.0 298.25 -23.0 -0.00081204
WGS-60 6378165.0 298.3 -28.0 0.00480795
WGS-66 6378145.0 298.25 -8.0 -0.00081204
WGS-72 6378135.0 298.26 2.0 0.0003121057
WGS-84 6378137.0 298.257223563 0.0 0.0

Table 2: 1. Reference Ellipsoid, 2. semi-major axis (a in m), 3. reciprocal flattening (1/f ),


4. ∆a in m, 5. ∆f (∗104 ). ∆-parameters are with respect to WGS-84. Source: Defense
Mapping Agency. 1987b. DMA Technical Report: Supplement to Department of Defense
World Geodetic System 1984 Technical Report. Part I and II. Washington, DC: Defense
Mapping Agency

By defining the ellipsoidal surface as an equipotential surface, originating from a reference


gravity potential U , h can be expressed as a function of U as we discussed before using W
and H. N can be expressed as a function of the potential difference
T =W −U (146)
N = f (W − U ) = f (T ), (147)
where T is called the disturbing potential. A simple approximation of N is N = T /γ,
where γ is normal gravity. N can thus be defined in a physical sense as a functional of the
disturbing potential T . The determination of N and the broader question of approximating
T will be treated in the course ENGO423: Geodesy. Equation 145 is of particular interest in
satellite positioning where the ellipsoidal height h can be determined with high accuracy.
To determine the orthometric height H, which is needed in many applications, N has to
be determined with comparable accuracy. The determination of N is therefore one of the
central problems of physical geodesy. Physically meaningful heights requires both accurate
geometric and accurate physical measurements of heights and gravity. The small angle
θ between the normal to the ellipsoid and the plumbline is called the deflection of the
vertical, see Figure 31 for a geometrical explanation. It is usually split into a North-South
component ξ and a East-West component η, so that
θ2 = ξ 2 + η 2 . (148)

C Alexander Braun (2006-2010)


70 ENGO421: COORDINATE SYSTEMS
The two components describe the difference in curvature between the geoid and the ellip-
soid. They can therefore be related to the first gradients of the disturbing potential with
respect to Φand Λ. In a first approximation we can derive
1 δT
ξ= (149)
ḡ RE δΦ
1 δT
η= , (150)
ḡ RE cos Φ δΛ
where ḡ is the mean gravity of the Earth and RE is the mean radius of the Earth.
To connect a Geodetic reference system (GRS) to geodetic measurements, the origin and
the orientation of the attached coordinate frame need to be defined. This is done by
satellite geodetic methods and has resulted in the establishment of Geodetic Reference
Frames, such as WGS-84 and ITRF. Both of them use GRS-1980 to define the ellipsoidal
parameters.

5.2 Basic Ellipsoidal Geometry


A rotationally symmetric ellipsoid is generated by rotating an ellipse around its semi-minor
axis b. Two geometric parameters, the semi-major axis a and the semi-minor axis b deter-
mine the rotational ellipsoid entirely. Obviously, meridional sections of the ellipsoid are
equal because they are based on the same ellipse. This characteristic is called rotational
symmetry (Figure 33).

Figure 33: Ellipsoidal geometry and semi-major and semi-minor axes.

Several other parameters can be used to describe the ellipsoidal geometry, but only two
are necessary to describe the ellipsoid completely. The following relations allow for the
computation of all other parameters if two are known.
a−b
f= flattening (about 1:298 for the Earth) (151)
a
a2 − b 2
e2 = first eccentricity (152)
a2
a2 − b 2
e02 = second eccentricity (153)
b2
C Alexander Braun (2006-2010)
71 ENGO421: COORDINATE SYSTEMS
E 2 = a2 − b2 linear eccentricity (154)
(1 − e2 )(1 + e02 ) = 1 (155)
e02
e2 = = 2f − f 2 (156)
1 + e02
02 e2 2f − f 2
e = = (157)
1 − e2 1 − (2f − f 2 )
√ 1
f = 1 − 1 − e2 = 1 − √ (158)
1 + e02
b √ 1 e
= 1 − e2 = √ = 0 =1−f (159)
a 1+e 02 e
a 2
a √ b 02 a
=√ = a 1 + e 02 =
2
= b(1 + e ) = (160)
b 1 − e2 1−e 1−f

The flattening f and the first and the second eccentricities can be expanded into power
series with respect to e2 or e02 :

e2 e4 e6 5e8
f= + + + + +... (161)
2 8 16 128
e02 3e04 5e06 35e08
or = − + − ... (162)
2 8 16 128
e2 = e02 − e04 + e06 − e08 ... (163)
e02 = e2 + e4 + e6 + e8 + ... (164)

In the following example, these relations are used.



f = 1 − 1 − e2 = 1 − (1 − e2 )1/2 (165)

Since e2 << 1 a binomial series can be employed. Its general form is:

q(q − 1) 2 q(q − 1)...(q − k + 1) k


(1 + x)q = 1 + qx + x + ... + x (166)
2! k!
With q = 1/2, the term (1 − x)1/2 therefore becomes
1 1 1 5 4
(1 − x)1/2 = 1 − x − x2 − x3 − x − ... (167)
2 8 16 128
substituting x = e2 (168)
e2 e4 e6 5e8
f =1−1+ + + + + .... (169)
2 8 16 128

5.3 Coordinates on the Ellipsoid - Geodetic Coordinate Systems


In chapter 2 we used the gravity vector and the spin axis of the Earth as references to
establish the coordinate frames LA and CT . However, we learned that the knowledge
of the gravity field is mandatory and would like to simplify the coordinate system by
introducing a mathematically defined reference surface which can be used without any

C Alexander Braun (2006-2010)


72 ENGO421: COORDINATE SYSTEMS
Physical world Mathematical/Geometrical World
meaningful exact
geoid ellipsoid
gravity ~g normal gravity ~γ
gravity potential W normal potential U
astronomical coordinates Λ, Φ geodetic coordinates λ, φ
LA LG
CT G
orthometric height H ellipsoidal height h
potato pomegranate, onion

Table 3: Relations between the physical/geoid and the geometrical/ellipsoidal world.

additional knowledge of the natural coordinates to set-up a coordinate frame. A point


on the ellipsoidal surface can be defined either by curvilinear coordinates (φ, λ, h) or by
Cartesian coordinates (x, y, z).
The use of curvilinear coordinates appeals more to our intuition and has for a long time
been the standard method of representing points in space. Three-dimensional Cartesian
coordinates have gained in importance after satellites measurements, e.g. Global Naviga-
tion Satellite Systems (GNSS), became a major component of geodetic operations. The
curvilinear representation makes use of the fact that the point has to be on the surface of
the ellipsoid, and position can thus be defined by two coordinates only. This will be the ap-
proach taken in this chapter. In the Cartesian representation, three coordinates are needed
because the surface is embedded in three-dimensional space. In practice, this difference is
of little significance because measurements take place in 3-D space and are not confined
to the surface of an ellipsoid. Thus, the curvilinear as well as the Cartesian representation
requires three numbers to define a position in space. Therefore, a third coordinate, the el-
lipsoidal height h, has to be defined to represent measurements by curvilinear coordinates.
h does now take the place of H in previously established coordinate frames LA and CT .
Since the ellipsoid is a close approximation of the geoid, the concepts outlined in Chapter 2
can easily be transferred to the ellipsoid. In general, the following parameters can be seen
as equivalent. This table must be taken as guidelines only to obtain a better understanding,
the exact relations will be introduced in the following.

Thus, natural coordinates and the local astronomic system have their equivalents in ellip-
soidal geometry. The main simplification is due to the rotational symmetry of the biaxial
ellipsoid. Thus, in contrast to the situation on the geoid, each meridian plane has the
same geometric properties. A meridional plane of the ellipsoid is always an ellipse of the
same dimensions (analogy to slicing a pomegranate or onion). An astronomical meridian
section always looks different depending on which longitude it is (analogy to slicing a
potato). The longitude λis the same for all curvilinear systems on the ellipsoid and it is
therefore possible to restrict the discussion to an arbitrary meridian plane. In this plane,
the x and y axes in the equatorial plane can be replaced by the p-axis which is defined
as the intersection of the meridian plane with the equatorial plane (Figure 34). Thus, a
point in the meridian plane is either defined by the latitude or by the coordinates p and z.
Different definitions of the latitude are possible and three of them will be discussed later
in this chapter.

C Alexander Braun (2006-2010)


73 ENGO421: COORDINATE SYSTEMS
Figure 34: Geodetic coordinates (λ, φ) on the ellipsoid.

Once p and z have been defined, the x, y-coordinates in a suitably defined Cartesian sys-
tem, as for instance the CT-system, can be obtained from simple trigonometry

x = p cos λ (170)
y = p sin λ. (171)

Thus, the transformation into a three-dimensional Cartesian system is straightforward, if


λ is known.

Geodetic Coordinates

In Chapter 2, astronomical or natural coordinates Φ, Λhave been defined. They are given
by the direction of the gravity vector in the CT-system. This vector is normal to the equipo-
tential surface passing through the point under consideration. Geodetic coordinates φ, λ
are defined in a similar way. They are given by the direction of the normal to the ellipsoid
in the CT-system. This vector corresponds to the normal gravity vector ~γ if the ellipsoidal
surface is the equipotential surface defined by the reference potential U . Geodetic coordi-
nates are important because they are close approximations of the ’observables’ Φ, Λ. The
local geodetic system is defined in the same way as the local astronomic system, except
that the reference figure is the ellipsoid, not the geoid (Figure 35).
The parameter of the LG-system are:

C Alexander Braun (2006-2010)


74 ENGO421: COORDINATE SYSTEMS
Figure 35: Local Geodetic System (λ, φ) on the ellipsoid.

• Local Geodetic Frame (LG)


• Origin: At observers point P
• Primary axis (z): Orthogonal to ellipsoid at P
• Secondary axis (x): Tangent to geodetic meridian pointing north
• Tertiary axis (y): complete a left-handed system (LHS)

Note that the z-axis coincides with the ellipsoidal normal at P and points in the direction
opposite to the normal gravity vector ~γ at P , if the normal ellipsoid is used. The (x, y)-
plane coincides with the plane tangent to the ellipsoid at P and normal to ~γ . The x-axis
lies in the meridian plane and is oriented towards north. The y-axis points eastward. This
makes the local geodetic system (LG) a left-handed system. The vector ~n normal to the
ellipsoid is  
cos φ cos λ
n~CT =  cos φ sin λ  (172)
sin φ
in complete analogy to equation (58). As previously, φ and λ define the direction of the
surface normal in space. Geodetic coordinates φ, λ are very close to astronomic coordinates
Φ, Λ. The difference does usually not exceed ±1 arc-minute and more typically is about

C Alexander Braun (2006-2010)


75 ENGO421: COORDINATE SYSTEMS
15 arc-seconds or less. As described before, the difference between the astronomic and
geodetic latitude and longitude is described by the deflections of the vertical.
ξ = Φ − φ N-S vertical deflections (173)
η = (Λ − λ) cos φ E-W vertical deflections (174)
Here they are geometrically interpreted as changes in direction between the astronomic
and the geodetic coordinate systems. The north-south deflection ξ can be shown along the
meridians as the E-W component disappears (Figure 36). It is important to realize that
vertical deflections describe the curvature difference between the geoid and the ellipsoid,
or as the equations outlines, the geometric difference of the angles Φ − φ and Λ − λ,
respectively. On the other hand, vertical deflections also reflect the different orientation
of the two physical parameters ~g and ~γ . They are fundamental to describe the difference
between geoid and ellipsoid. The parameters φ, λ can also be considered as coordinates on

Figure 36: North-South vertical deflection ξ on the meridian.

the ellipsoidal surface. The parallels and meridians then fulfill the following conditions:
parallels: φ = const (175)
meridians : λ = const (176)
It should be noted that rotational symmetry is true for geodetic coordinates and does not
hold for astronomic coordinates as the latter refer to the geoid which is not rotationally
symmetric. In the literature, geodetic coordinates are frequently called geographical co-
ordinates. In geodesy, the term geodetic coordinates is preferred, but it is important to
realize that the term geographical coordinates is identical when communicating with ge-
ographers. o Comparable to the relation between LA and CT , we will now introduce the

C Alexander Braun (2006-2010)


76 ENGO421: COORDINATE SYSTEMS
global equivalent of CT for the ellipsoid, the Global Geodetic System (G). Global Geodetic
coordinate systems may refer to ellipsoids of different dimensions and different origins. Al-
though all of them are ’close’ to the CT-system in some sense, they are not always identical
to it. Such systems are called Global Geodetic Systems and are defined by

• Global Geodetic System (G)


• Origin: Geometrical centre of the particular ellipsoid used
• Primary axis (z): Rotation axis b of the ellipsoid
• Secondary axis (x): Greenwich Meridian
• Tertiary axis (y): Complete a right-handed system (RHS)

Global Geodetic Systems are widely used in geodesy to link the local coordinate systems
to a global reference. They may differ in origin, orientation, and ellipsoid dimensions
from the CT -system. The reasons for these differences are historical and major efforts
are currently made to refer all coordinates to a unique global reference system such as
the CT -system. The practical implementation of such a system via GNSS satellites will be
discussed later. In order to allow for easier transformations between the systems and also
to better understand the relations between the systems, a number of additional parameters
must be introduced. They will later allow us to move from one system to another without
having to develop complicated rotations and translations.

The Reduced Latitude

In order to develop mathematical tools to transform between different coordinate systems,


a number of different latitude must be defined. As seen before, the astronomical latitude
Φ is not defined with respect to the Earth’s center of mass, but is the angle between the
equator and the direction of the gravity vector ~g extended to the equator. The geodetic
latitude φ is defined in a similar way using the direction of the normal gravity vector ~γ .
Again, this vector does not necessarily point to the center of the ellipse. For mathematical
derivations, it is better to define the latitude angle at the Earth’ centre of mass or the centre
of the ellipse. This is achieved with the concept of the reduced latitude β. The reduced
latitude β of the point P on the surface of the ellipsoid with semi-axes a and b is given
shown in Figure 37. It shows a section of the ellipsoid along the meridian plane. The
horizontal distance p of the point P from the z-axis is the radius of the parallel
q
p= x2 + y 2 (177)

The parametric representation of the meridian ellipse can be derived from Figure 37. From
the circle with radius a we get
p0 = p = a cos β (178)
z 0 = a sin β (179)
From the circle with radius b we obtain
b
z = z 0 = b sin β (180)
a
C Alexander Braun (2006-2010)
77 ENGO421: COORDINATE SYSTEMS
Figure 37: Reduced latitude β and relation to the geodetic latitude φ.

In summary, the reduced latitude can be obtained by projecting the ellipse on the concen-
tric circle with radius a of the semi-major axis. It is convenient to use in derivations and
is important in the theory of ellipsoidal mapping. Note that β always is measured at the
centre of the ellipse and does not move along the axes a or b as φ does. While this is the
desired characteristic, it is not convenient as the projection of the point P to P 0 is required.
The next latitude definition will overcome this and P remains on the ellipsoid.

The Geocentric Latitude

The concept of the geocentric latitude is depicted in Figure 38. The relations between the
geocentric latitude γ and the coordinates p and z can be easily derived. Note, γ is not to
be confused with the normal gravity vector ~γ .
q
r = p2 + z 2 (181)
p = r cos γ (182)
z = r sin γ (183)

The geocentric latitude is very convenient and frequently used for computations on the
ellipsoid surface.

C Alexander Braun (2006-2010)


78 ENGO421: COORDINATE SYSTEMS
Figure 38: Geocentric latitude γ.

Relations Between Φ, φ, β, γ

By now, we have four different latitude definitions in this course. As with the coordinate
systems, we need all of them to establish the desired coordinate system for observations,
computations, and transformations. One of the latitudes can be measured, it is the as-
tronomical latitude Φ. All the others are definitions which can be related to Φ. As we
have seen before, the vertical deflections connect the astronomical and geodetic latitude.
Now we will establish the relation between the geodetic, the reduced and the geocentric
latitude. This gives us a full set of relations which allows for transformations from any
latitude to any other latitude.

Relation between φand β : Differentiating equations (178, 179) with respect to βand
using the slope of the ellipsoidal tangent allows for the derivation of relations between φ
and β. We skip the intermediate derivations and only present the final equation:
b
tan β = tan φ (184)
a
The maximum difference between the geodetic latitude φ and the reduced latitude β can
be estimated from a Taylor series expansion and is 50 .5 arc-minutes at latitude φ = 45o .

Relation between φ and γ : The relation can be expressed using the previous equations
and results in:
b2
tan γ = tan φ (185)
a2

C Alexander Braun (2006-2010)


79 ENGO421: COORDINATE SYSTEMS
The difference between geodetic latitude φ and geocentric latitude γ can be obtained using
a similar series expansions as before. The difference (φ − γ) between geodetic and geo-
centric latitude is twice as much as the difference (φ − β) between geodetic and reduced
latitude. Its maximum is 110 arc-minutes and it occurs at latitude φ = 45o .

Latitudes as Functions of p and z : Equations (178, 179) give the parametric represen-
tation of Cartesian coordinates p and z as a function of β, and equations (182, 183) give it
as a function of γ. Here, we derive the parametric representation of p and z will be derived
as a function of φ. This will be done by expressing sin β and cos β in terms of the geodetic
latitude φ. After a number of derivations and a substitution of

W 2 = 1 − e2 sin2 φ (186)

we arrive at the following formulas relating the Cartesian coordinates p and z to the ellip-
soidal parameters and the geodetic latitude.
a
p= cos φ (187)
W
a(1 − e2 )
z= sin φ (188)
W
a q
r= 1 + (e4 − 2e2 ) sin2 φ (189)
W
By using these formulas, geodetic coordinates on the surface of the ellipsoid can now
be expressed by three-dimensional Cartesian coordinates. This transformation is used to
represent the origin of the local geodetic system in global coordinates. The following
three formulas summarize the results for all three latitude parameterizations using the
substitution N = a/W , which is the prime vertical of curvature. The formulas show a
very similar structure but differ in the length parameter which pre-multiplies the direction
vector and in the rate of change for the z-coordinate. Essentially, both changes describe a
variable radius of curvature for the surface.
 CT  
x cos φ cos λ
Geodetic latitude  y  = N  cos φ sin λ  (190)
z (1 − e2 ) sin φ
 CT  
x cos β cos λ
 cos β sin λ 
Reduced latitude  y  =a  √  (191)
z 1 − e2 sin β
 CT  
x cos γ cos λ
Geocentric latitude  y  = r  cos γ sin λ  (192)
z sin γ

5.4 Transformation Between Cartesian and Curvilinear Geodetic Co-


ordinates
The transformation between Cartesian and curvilinear coordinates plays an important role
in geodesy since terrestrial measurements are usually expressed in curvilinear coordinates

C Alexander Braun (2006-2010)


80 ENGO421: COORDINATE SYSTEMS
φ, λ, h, while satellite observations are usually referred to a Cartesian x, y, z-system. Curvi-
linear coordinates make more sense to us as we can relate our position easier than with
Cartesian coordinates. Assuming that the point P in Figure 39 has known curvilinear
geodetic coordinates φ, λ, h and that the Cartesian coordinates x, y, z of P are to be deter-
mined, i.e
φ, λ, h =⇒ x, y, z
Based on Figure 39, the distance vector r~P can be expressed by

r~P = r~Q + h ~n (193)


 
cos φ cos λ
with r~Q = N  cos φ sin λ  (194)
(1 − e2 ) sin φ

The normal vector to the ellipsoid ~n is given by

Figure 39: Cartesian and Curvilinear Coordinates


.

 
cos φ cos λ
~n = cos φ sin λ 
 (195)
sin φ

C Alexander Braun (2006-2010)


81 ENGO421: COORDINATE SYSTEMS
By adding the vectors r~Q and ~n an equation for r~P can be derived, this completes the
transformation of curvilinear coordinates on the right-hand-side to Cartesian coordinates
on the left-hand-side.
   
x (N + h) cos φ cos λ
r~P =  y  =  (N + h) cos φ sin λ  (196)
z ((1 − e2 ) N + h) sin φ

As this transformation is straightforward, it does not require significant computations to


derive the Cartesian coordinates, but this is different for the inverse case. We know like
to transform known Cartesian coordinates x, y, z to the equivalent curvilinear coordinates
φ, λ, h,
x, y, z =⇒ φ, λ, h.
Unfortunately, there is no linear relationship between Cartesian and curvilinear coordi-
nates. In order to express curvilinear coordinates in terms of Cartesian coordinates either
a 4th order equation has to be solved analytically or an iterative solution has to be applied.
The iterative solution will be presented here because it provides insight into the method
and results in a simple algorithm. Analytical solutions have been published but are not
subject of this course. From the equation for r~P the longitude λ can be obtained directly
y
tan λ = (197)
x
However, the other two coordinates φ and h must be iterated because the other two equa-
tions contain both of them. Using the first two equations gives the relation of p and h, φ.
q
p= x2 + y 2 = (N + h) cos φ (198)

After solving for g, this gives the first equation for the iterative process
p
h= − N. (199)
cos φ
The last equation in r~P gives
z = (N − e2 N + h) sin φ, (200)
which can be used to obtain the corresponding equation for φ. Dividing zP by p leads to
z N − e2 N + h N
= tan φ = (1 − e2 ) tan φ (201)
p N +h N +h
which can be solved for tan φ to get the third equation of the transformation process.
z N −1
tan φ = (1 − e2 ) (202)
p N +h
In the iterative equations (199, 202), the parameters z and p do not change during the
iteration, but φ, N and h must be iterated as they depend on each other in the equations.
Looking at equation (201), we realize that the term in the brackets will simplify to (1 − e 2 )
if h = 0 and N cancels out. Therefore, the iteration starts by setting h = 0, which results
in an equation which does not contain any phi or h on the right-hand-side. The procedure
to derive φ, λ, h from x, y, z is therefore:

C Alexander Braun (2006-2010)


82 ENGO421: COORDINATE SYSTEMS
1. Compute λfrom (197)
2. Determine initial estimates for h (here, h = 0) and derive φ0 from
z
tan φ0 = (1 − e2 )−1 (203)
p

3. Use φ0 to derive the first N0 from


a a
N0 = =q (204)
W 1 − e2 sin2 φ0

4. Derive h1 from
p
h1 = − N0 (205)
cos φ0
5. Finally, use N0 and h1 in
z N0
tan φ1 = (1 − e2 )−1 (206)
p N0 + h1

to derive the next φ1 .

Go back to step 3. and repeat the iteration till convergence is achieved. In general, the
convergence criterion should be set for either h or N and should correspond to the reso-
lution of the initially given Cartesian x, y, z-coordinates, e.g. if cm is the least significant
digit for x, y, z, the iteration should be stopped at the cm-level. It should be noted that this
algorithm has a singularity at p = 0, i.e. for points on the spin axis, as p appears in the
denominator. Since this corresponds to x = 0 and y = 0, this case can be easily eliminated
by checking the input coordinates.
This concludes the definition of geodetic coordinates and coordinate systems. We now
understand the relation between astronomical coordinate systems and geodetic coordi-
nate systems. The next step will be the transformation between such systems using the
parameters defined in this chapter.

C Alexander Braun (2006-2010)


83 ENGO421: COORDINATE SYSTEMS
6 Coordinate Transformations
In the previous chapters, different coordinate systems have been developed, they have
different purposes and applications. The systems can be distinguished in two groups, ter-
restrial and celestial systems. In addition, the terrestrial can be divided into astronomical
frames and geodetic frames, depending on the reference surface and axes used in the def-
inition. As all measurements in a physical environment Earth take place in astronomical
frames, the transformation into other systems must be defined and will be discussed here.
Here, the transformation between the terrestrial coordinate systems will be discussed, the
transformation between terrestrial and celestial coordinate systems is subject to Earth ro-
tation, polar motion, precession and nutation, all time dependent processes which rotate
the coordinate frames with time. These processes will be covered in the course ENGO423:
Geodesy. Only then, the coordinate transformation can include time dependent transfor-
mation parameters. Here, only the static cases will be explained.
Coordinates of points are usually determined in a system which is convenient for the task
at hand. It can be local, regional, or global, and it may be Cartesian, polar, or curvilinear
depending on the specific problem. Since geodesy deals with the figure of the Earth, global
coordinates are most appropriate also with respect to exchange of information and data.
If coordinates are established by measurement on the Earth’s surface, curvilinear coordi-
nates are usually most appropriate as we understand their meaning easier than Cartesian
coordinates. Additionally, they have the advantage that the distinction between horizontal
coordinates and height coordinates is very intuitive. ON the other hand, satellites or other
extraterrestrial methods for positioning on the Earth’s surface mostly employ 3-D Carte-
sian coordinates. The transformation between Cartesian and curvilinear coordinates on
the ellipsoid has already been treated in the last chapter. It does not present a problem
as long as both sets of coordinates refer to the same coordinate system, i.e. including the
same origin, the same axes orientation, and the same scale. In practice, this is often not
the case. Positions derived from satellite observations are essentially in a CT system when
taking the accuracy of the measurement into account. Positions derived from terrestrial
measurements are either in a local system (LA) or refer to a global system (G) that has
small translations and rotations with respect to the CT system and may also have differ-
ences in scale. In order to transform the coordinates of such a system into the CT system
is called the datum problem. The datum problem describes the problem of transforming
all coordinates and measurements to the same unified datum.
Before satellite methods with sufficient accuracy became available, the datum problem
was very difficult to solve as there were not many observations available which allowed
geodesists to connect networks across continents and globally. Today, GNSS observations
are tied into a well-maintained CT frame (particularly the International Terrestrial Refer-
ence Frame - ITRF). Today, the problem can essentially be reduced to a series of coordinate
transformation in Cartesian coordinates. These transformations will be given in this chap-
ter. Further, the equivalent formulas for curvilinear transformations exist, called differen-
tial formulas, but these will not be discussed in this course. Finally, some of the remaining
problems relating to datum transformations will be outlined.

C Alexander Braun (2006-2010)


84 ENGO421: COORDINATE SYSTEMS
6.1 Transformation Between Systems with Different Origins and Ori-
entations
As outlined before, an best fitting ellipsoid is developed by minimizing the squares of the
geoid undulations N between geoid and ellipsoid. Depending on the fact if a local or global
fit is performed, the origin and the axes a and b can be different. This leads to different
origins between G and CT . The transformation between coordinate systems with different
origins and orientations is frequently needed in geodesy because

• coordinate systems used in classical triangulation refer to an arbitrary fundamental


point and have orientations which are derived from astronomical observations
• coordinate systems underlying satellite positioning refer to the Earth’s mass centre
and derive their orientation from the dynamical system of satellite orbits.

Thus, differences in origin and orientation can occur between these two types of systems
as well as between different triangulation systems and different satellite positioning sys-
tems. In the following, the transformation between a Global Geodetic System (G) and the
Conventional Terrestrial System (CT ) will be discussed. The G-system is typically a sys-
tem derived from terrestrial triangulations, while the CT -system is typically derived from
satellite observations. The resulting transformation formulas can be applied to other cases
with only minor modifications. Herein, rotation and reflection matrices and translation
vectors will be used frequently to transform between coordinate systems. Please refer to
chapter 3 for a discussion of the rotation matrices.

Figure 40: Relation between the Global Geodetic System G and the Conventional Terres-
trial System CT .

C Alexander Braun (2006-2010)


85 ENGO421: COORDINATE SYSTEMS
The relation between the two systems is shown in Figure 40. The origin of the G-system
is shifted with respect to the origin of the CT -system by the vector and the axes of the
G-system are rotated with respect to the CT -system by 1 , 2 , 3 about the x, y, z-axis, re-
spectively. The transformation of the G-system into the CT -system,
G =⇒ CT
is often referred to as the datum problem and consists of the following steps:

• Compute Cartesian coordinates x, y, z of P in the G-system using equation (196)

• Rotate the r~PG -vector into the CT -system


• Shift the origin of the G-system into the CT -system using r~0 .

The transformation formula becomes

r~CT = r~0 + R1 (1 ) R2 (2 ) R3 (3 ) r~PG (207)


where  
x0
~r = y0 
 (208)
z0
and r~PG is given by equation (196). Typically, the rotation angles i are very small and the
small angle approximation can be applied. Equation (207) shows that the transformation
of a G-system into the CT -system or, in other words, the solution of the datum problem,
depends on 6 parameters: 3 translational parameters and 3 rotational parameters. In
order to determine these six transformation parameters, the coordinates of at least two
points must be known in both coordinate systems. If a scale factor exists between the two
systems, then a seventh parameter has to be added. A scale factor is likely if different
instruments have been used to determine the scale in each system; as for instance invar
wires to determine the baselines of a terrestrial triangulation, and time measurements
and thus the speed of light, to determine the scale of a satellite network. The inverse
transformation of the CT -system into the G-system, i.e.
CT =⇒ G
can be obtained by inverting equation (207) as discussed in the chapter on rotation matri-
ces.
r~PG = R3 (−3 ) R2 (−2 ) R1 (−1 ) (r~CT − r~0 ) (209)

An example for this transformation is the conversion from the North American Datum
(N AD1927) to the CT -system. As the rotation angles are all very close to zero, the trans-
formation equation simplifies to
~ = r~0 + r N AD1927
r CT ~ (210)
with r~0 = (−15m, 165m, 175m). Typical numerical values for the transformation of
N AD1927 to the CT -system are:

x = −15m ± 2m = 0 ± 0.”15
y = 165m ± 2m = 0 ± 0.”15
z = 175m ± 2m = 0 ± 0.”15

C Alexander Braun (2006-2010)


86 ENGO421: COORDINATE SYSTEMS
The reference ellipsoid underlying the N AD1927 system is the Clarke (1866) ellipsoid
with the values given in Table 2. Note that this simplification is not necessarily the case
for other datums as for instance the European datum or the Australian datum, where
small rotations must be taken into account as well. As noted above, only a translation
of the origin is required to transform between the two global terrestrial systems, the next
section will show that the transformation between the global and the local systems is more
complicated.

6.2 Transformations Between Local and Global Systems


Terrestrial geodetic measurements are usually made in the LA-system, e.g. astronomic
azimuth determination done in the lab assignment. Coordinates are usually computed in
a global system (G), which allows for exchange of coordinates between different clients,
agencies or countries. The transformation between these two systems will be discussed in
this section, using the LG-frame as an intermediate system. The steps to be discussed are
therefore
LA =⇒ LG =⇒ G
Consider first the transformation of the LA-system into the LG-system. As shown in Figure
41, both systems have the same origin and the directions of their axes differ by small
amounts only, which is expressed by the deflections of the vertical. The transformation
matrix will therefore be composed of a sequence of small angle rotations. These angles
will now be derived. The z-axis of LA-system coincides with the vector n~LA normal to
the geoid while the z-axis of LG-system has the direction of the vector n~LG normal to the
ellipsoid. The spatial angle θ between them defines the total deflection of the vertical.
Since the deflection of the vertical θ is a spatial angle, it is convenient to decompose it into
mutually orthogonal components. The north-south component in the (x, z)-plane defined
by the astronomic meridian is denoted by ξ, and the east-west component in the (y, z)-
plane defined by the prime vertical is denoted by η.
Figure 42 shows a unit sphere around P1 and the spherical coordinates defining the differ-
ent vectors emanating from P1 . The latitude and longitude difference between the points
where z LA and z LG pierce the unit sphere are the components ξ and η of the total deflection
of the vertical θ. They are given by

ξ = Φ − φ N-S vertical deflections (211)


η = (Λ − λ) cos φ E-W vertical deflections (212)

By projecting η into the horizon plane (x, y-plane), the difference between the astronomic
azimuth A and the geodetic azimuth α is obtained by

∆A = A − α = (Λ − λ) sin φ (213)

This can also be expressed as


∆A = η tan φ. (214)
The formula transforming the LA-system into the LG-system becomes

r~LG = R3 (∆A) R2 (−ξ) R1 (η) r~LA (215)

C Alexander Braun (2006-2010)


87 ENGO421: COORDINATE SYSTEMS
Figure 41: Relation between the Global Geodetic System G and the Conventional Terres-
trial System CT .

As the vertical deflections are small quantities, the largest values occur in high mountains
and do not exceed 10 (seldom larger than 15”), the transformation formula can exploit the
small angle property and can be approximated by
 
1 ∆A ξ
~
r = −∆A 1 η  r~LA
LG  (216)
−ξ −η 1

In Figure 42, it has been assumed that the point P1 is on the geoid (or h = N ). This is
usually not the case. If it is at the orthometric height H above the geoid, two approaches
are possible. Either the deflection of the vertical is defined at P1 , i.e. at the equipotential
surface passing through P1 , or it is downward continued to the geoid along the curved
plumbline. This difference in the definition of the deflections of the vertical plays a major
role in the formulation of the boundary value problems (BVP) of geodesy and its discussion
will therefore be continued in the course ENGO423: Geodesy. The inverse transformation,
i.e. the transformation of LG into LA can be easily obtained by inverting equation (215)

C Alexander Braun (2006-2010)


88 ENGO421: COORDINATE SYSTEMS
Figure 42: Relation between the deflections of the vertical and the unit sphere around P1 .

using the properties of rotational matrices.

r~LA = R1 (−η) R2 (ξ) R3 (−∆A) r~LG (217)

The transformation of the LA-system into the G-system can be considered as a sequence
of two transformations: first the LA-system is transformed into the LG-system and then
the LG-system is transformed into the G-system. The first transformation has already been
examined, thus only the second one will be discussed now.
LG =⇒ G
The transformation of the LG-system into the G-system consists of two steps:

• Rotate the axes of the LG-system into the G-system


• Shift the origin of the LG-system into the origin of the G-system.

To rotate the LG-system into the G-system, the following operations have to be performed
as outlined in Figure 41:

1. Make the LG-system right-handed - P2


2. Align the z-axes - R2 (90o − φ)

C Alexander Braun (2006-2010)


89 ENGO421: COORDINATE SYSTEMS
3. Bring the y-axes together - R3 (180o − λ)
4. Shift the origin of LG-system to the origin of the G-system.

The first three steps, i.e. the rotations, are done by

∆r~G = r~2G − r~1G = R3 (π − λ) R2 (π/2 − φ) P2 r12


~LG , (218)
which makes the axes of the two systems parallel. The last step, the origin translation, is
done by bringing r~1G the right-hand side. Thus,

r~2G = r~1G + R3 (π − λ) R2 (π/2 − φ) P2 r12


~LG (219)

where r~1G is given by equation (196). The formula transforming the LA-system into the
G-system can be obtained by substituting equation (215) into (219). It takes the form

r~2G = r~1G + R3 (π − λ) R2 (π/2 − φ) P2 R3 (∆A) R2 (−ξ) R1 (η) r12


~LA (220)
This formula can be simplified by using astronomic latitude and longitude instead of
geodetic quantities. It becomes

r~2G = r~1G + R3 (π − Λ) R2 (π/2 − Φ) P2 r~12


LA
, (221)
where we obtain Λ and Φ from equations (211, 212).
Φ = φ+ξ (222)
Λ = λ + η secφ (223)
In can be shown that formulas (220) and (221) are equivalent. Note that aligning the
LA-frame to the G-frame requires only Φ and Λ. The use of φ, λ, ξ and η is obviously
equivalent. The transformation of one system into the other requires another parameter,
h or H + N . Note also that the origin of the G-system and the orientation of its axes may
not coincide with those of the CT -system. In that case, the transformation described in
the previous chapter must be applied as well.

Lab Assignment 3 Transformations: For lab 3, the coordinates given in the G-system
must be transformed into the LG-system in order to allow the SLR station operator to
orient the Laser to the satellite with correct azimuth A and zenith angle z. The following
transformation is required.

r~2G = r~1G + R3 (π − λ) R2 (π/2 − φ) P2 r12~LG (224)


   
x (N + h) cos φ cos λ
r~1G =  y  =  (N + h) cos φ sin λ  (225)
z ((1 − e2 ) N + h) sin φ
r~LG = P R (φ − π/2) R (λ − π) (r~G − r~G )
12 2 2 3 2 1 (226)
q
2 2 2
distance d12 = ∆xLG12 + ∆y12
LG LG
+ ∆z12 (227)
azimuth α12 = tan−1 (∆y12LG
/∆xLG
12 ) (228)
−1 LG
zenith angle z12 = cos (∆z12 /d12 ) (229)

C Alexander Braun (2006-2010)


90 ENGO421: COORDINATE SYSTEMS
6.3 The Datum Problem Today
In the classical datum problem the differences in rotation, translation, and scale of two G-
systems was determined from measurements. One of these systems is usually considered
as a close approximation of a CT -system. Satellite techniques have changed the approach
to the datum problem in a significant way. They are not only a means of establishing
a very precise CT -system by measurement, but they also provide, for the first time, the
measurement precision necessary to monitor changes of this system in time and space.
Further, they allow for estimating velocity vectors to point clusters in different regions of
the world. This has created a new field of studies called geodynamics, which relates the
measured changes to geophysical processes inside and above the Earth. Results of these
studies are affecting geodetic operations more and more as new dynamic processes are
discovered which include mass transfer, loading problems, and deformations of the Earth
surface.
Although there are many applications in which it is appropriate to consider positions of
points as not changing in time, the number of applications which require the monitoring
of positions in time is steadily increasing. Thus, it is of increasing importance to establish
global Reference Frames (RF) which are well defined in time, or in other word, which have
known deformations over time. The two international agencies currently providing RF ser-
vices to the international user community have both been established by the International
Association of Geodesy (IAG). They are the International Earth Rotation Service (IERS)
which maintains the IERS Terrestrial Reference Frame (ITRF) and the Earth rotation pa-
rameters, and the International GNSS Service (IGS) which maintains the IGS frame.
Reference frames are typically defined by a global network of carefully selected stations.
Coordinates of these stations at a specific epoch are defined by extraterrestrial measure-
ments using the best Earth orientation model available. Then, they are supplemented by
a set of velocity vectors with respect to that epoch, which define the changes of the RF
with time. Since most of the changes are slow, it is usually possible to use the same refer-
ence epoch for a considerable period of time. Currently, a new ITRF is released every 1-2
years. All stations belonging to a global network participate in an observational program
which continuously monitors and updates the changes in the system (e.g. plate motions,
pot-glacial rebound). The establishment of such reference frames by satellite (SLR, GNSS,
DORIS) and VLBI methods and their relation to the physics of the Earth are subject to
geophysics. Examples of such satellite systems will be discussed in the course ENGO423:
Geodesy.
To use a global RF for smaller regions, a network of fiducial points has to be established.
They can be stations of the RF in the region or stations that are permanently observed
and can therefore be easily connected to the global network. Requirements for fiducial
stations are that they are in stable locations, that they are well distributed over the region,
and that their coordinates in the national datum are well known. Using these coordinates
and those derived for the RF system, a seven parameter transformation (3 translations,
3 rotations, 1 scale) is made between the two systems. The transformation parameters
are then used to transform all coordinates in the region into the RF system. Permanent
observations of the stations in the fiducial network will provide motion vectors for smaller
regions, not detectable by the widely spaced RF stations. The fiducial network can be
used for further densification and as an Active Control System (ACS). Such a system is
dynamic in nature and supports general geomatics applications and specifically geodetic
positioning, navigation, and spatial referencing.

C Alexander Braun (2006-2010)


91 ENGO421: COORDINATE SYSTEMS
Figure 43: Current network of permanently operating GNSS stations in Canada. Source:
Geodetic Survey Division, NRCan.

Geomatics Canada, through its Geodetic Survey Division, has been actively involved in
contributing to the international efforts (ITRF and IGS) and has pioneered the idea of
an Active Control System. Figure 43 shows the current Canadian Active Control System
(CACS) which, augmented by 18 globally distributed IGS stations, provides a continuous
geodetic reference for the territory of Canada. The accuracy of this reference is close to
that of the IGS which is estimated to have standard deviations of about 0.15 m in a global
geocentric frame.
A practical difficulty with transforming national networks to global RF by a seven pa-
rameter i transformation are inconsistencies in the national networks. The assumption
behind the transformation is that the national network is a rigid array of points with equal
accuracy everywhere. This is rarely the case, however. Triangulation and trilateration
networks have been observed over considerable periods of time with instrumentation of
differing accuracy and surveyors of different skills. Time changes in the coordinates over
time have not been monitored and have therefore led to distortions in the network. Sim-
ilarly, errors in the measurements and the unequal quality of the instrumentation has led
to inconsistencies in the network accuracy. Most of these effects will show up as local or
regional distortions in the networks which are obviously not eliminated by a seven pa-

C Alexander Braun (2006-2010)


92 ENGO421: COORDINATE SYSTEMS
rameter transformation. If enough network points in the region can be connected to the
fiducial network, these distortions can be modeled either by deterministic or by stochastic
methods.

6.4 Summary of Coordinate Systems and Transformations


Figure 44 shows the coordinate systems discussed so far subdivided into observational
and computational systems. The physical or geometric reference system is given for each
group. Also the mutual transformations including their parameters are indicated. This
figure outlines all systems discussed and the parameters which relate them. As none of
the systems includes time dependent parameters, we must develop coordinate systems
which include time variable parameters such as Earth rotation, this will be done in the
next course ENGO423: Geodesy.

7 Basics of Map Projections


Projecting a 3-dimensional space onto a 2-dimensional space (e.g. the surface coordinates
of your body onto the floor or the Rocky Mountains onto a 2-D map) is always associated
by distortions. No projection allows you to avoid all distortions. However, you can mini-
mize distortions and eliminate distortions in certain map properties. The reason why the
3-D to 2-D projection always creates distortions is the fact that the curvature of the 3-D
object is different than the 2-D projection. An example is the ellipsoid which has a posi-
tive curvature greater than zero while the curvature of any 2-D plane is equal to zero. It
is again the difference in curvature which creates problems in coordinate transformation.
Depending on the application, distortions may be acceptable, e.g. if you just like to learn
about the size of a country with respect to another country, you can take a 2-D map projec-
tion which preserves equal areas. The map projection used depends on the application and
the ultimate goal is the eliminate the distortions of the most important parameter. These
parameters describe properties of maps and include

• Equal area - the area on the 3-D surface is identical with the area on the 2-D surface.
Distortions are created in shape, angle, and scale.
• Conformal - the shape of structures are preserved in the 2-D plane. Distortions are
created in area and differential scale.
• Equidistant - The length of certain lines on the 3-D surface is identical to the line
on the 2-D projection plane. This does not hold for all lines between 2 points, but
selected ones only, hence, the equidistant property cannot be preserved for all lines
on a map. Distortions are created in shape, area and angle.

One important characteristic is that mapping from an ellipsoidal (or spherical) 3-D surface
onto a 2-D plane cannot be conformal and euqal area at the same time.

C Alexander Braun (2006-2010)


93 ENGO421: COORDINATE SYSTEMS
Figure 44: Coordinate Systems and Their Relationships

7.1 Direct and inverse geodetic problem


In geodesy and surveying, the most fundamental application is to derive the distance and
direction between two points. We distinguish between two problems, i) the direct geode-

C Alexander Braun (2006-2010)


94 ENGO421: COORDINATE SYSTEMS
tic problem, and ii) the inverse geodetic problem. The direct problem is to derive the
coordinates of point 2 given the coordinates of point 1 and the directions of point 2 wrt
point 1. The inverse problem is to derive the distance and directions between the two
points given both coordinates at points 1 and 2. The equations required for solving the
two problems are:
q
d12 = (x2 − x1 )2 + (y2 − y1 )2 + (z2 − z1 )2 (230)
y2 − y 1
α12 = tan−1 ( ) (231)
x2 − x 1
z2 − z 1
z12 = cos−1 ( ) (232)
d12
It does not matter if the coordinates of the points are given in Cartesian or curvilinear
coordinates as we can transform between them and keep sufficient accuracy.

7.2 Coordinate transformations in mapping


Mapping requires transformations to project 3-D coordinate onto a 2-D plane. The coor-
dinate system of the 2-D mapping plane can be defined as x pointing East and y pointing
North. In order to simplify the upcoming equations, the following quantities are intro-
duced:

t = tan φ (233)
a + b2
2
η2 = cos2 φ (234)
b2
W 2 = 1 − e2 sin2 φ (235)
a
N= (236)
W
x
xN = (237)
N

In the following we will introduce the mapping problem using ellipsoidal to mapping plane
transformations as these are the most relevant in geodesy and surveying. In the first step,
the 3-D coordinates are reduced to 2-D coordinates φ and λ for a given point P on the
ellipsoid, the height information disappears here as P is on the ellipsoid surface. In a
transverse Mercator projection, the 2-D coordinates xP and yP can be derived with the
following information about the difference of the longitude of P vs. the longitude of the
central meridian m.

∆λ = λP − λm (238)
n
X
yP = S(φQ ) + N A2i ∆λ2i (239)
i=1
n
X
xP = N A2i+1 ∆λ2i+1 (240)
i=0
with (241)
A1 = cos φ (242)

C Alexander Braun (2006-2010)


95 ENGO421: COORDINATE SYSTEMS
1
A2 = t cos2 φ (243)
2
1
A3 = cos3 φ(1 − t2 + η 2 ) (244)
6
1
A4 = t cos4 φ(5 − t2 + 9η 2 + 4η 4 ) (245)
24
1
A5 = cos5 φ(5 − 18t2 + t4 + 14η 2 − 58t2 η 2 ) (246)
120
1
A6 = t cos6 φ(61 − 58t2 + t4 + 270η 2 − 330t2 η 2 ) (247)
720
Once the coordinates xP , yP have been derived, the UTM coordinates can be obtained by
adding the scale factor k0 and the false easting x0 and false northing y0 , which is given as

x0 = 500000 m (f alse easting) (248)


y0 = 0 m if φ > 0 (249)
y0 = 10000 km if φ < 0 (f alse northing) (250)
k0 = 0.9996 : (scalef actor) (251)

False easting and northing have been defined to make all coordinates within one UTM
zone positive. The function S(φQ ) in the above equation is the Meridian Arc Length. The
point Q is located on the central meridian of λm and has the same latitude φ as the point
P . The meridian arc length between the equator and Q can then be derived.

yQ = S(φP ) (252)
S(φP ) = Aφ + B sin (2φ) + C cos (4φ) + D sin (6φ) + ....... (253)
with (254)
3 45 175
A = a(1 − e2 )(1 + e2 + e4 + e6 + ....) (255)
4 64 256
3 15 525 6
B = −a(1 − e2 )( e2 + e4 + e + ....) (256)
8 32 1024
15 4 105 6
C = a(1 − e2 )( e + e + ....) (257)
256 1024
35 6
D = −a(1 − e2 )( e + .....) (258)
3072
Based on the above series expansion, the accuracy of the Meridian Arc Length S is about
1.3cm for a distance of 10000km which is sufficient for the majority of applications.

ADDITIONAL INFORMATION ON MAP PROJECTIONS IS AVAILABLE FROM THE DOCU-


MENT “MAP PROJECTIONS” BY WHITTAL, BARRY AND MARTIN WHICH IS AVAILABLE
ON BB.

C Alexander Braun (2006-2010)


96 ENGO421: COORDINATE SYSTEMS

You might also like