Satellite Geodesy

by
Wolfgang Keller
Institute of Geodesy
University of Stuttgart
Copyright c 2004 by Wolfgang Keller
Contents
Chapter 1. Time Systems 3
1.1. Atomic Time 3
1.2. Dynamical Time 4
1.3. Sidereal and Universal Time 4
Chapter 2. Conventional Reference Systems 7
2.1. International Earth Rotation Service (IERS) 7
2.2. Celestial Reference System 7
2.3. Terrestrial Reference Systems 7
2.4. WGS84 8
2.5. Ellipsoidal Reference Systems 10
Chapter 3. Earth Rotation 13
3.1. Motion in Celestial System 13
3.2. Precession transformation 13
3.3. Nutation transformation 14
Chapter 4. Signal Propagation 17
4.1. Structure of the atmosphere 19
Chapter 5. Theory of Satellite Motion 23
5.1. Equation of Motion 23
5.2. Integration of Keplerian equation of motion 24
5.3. Disturbed Keplerian Motion 31
5.4. Variation of integration constants for disturbed orbit computation 33
5.5. Integration of the disturbing equations 36
Chapter 6. Numerical Orbit Integration 45
6.1. Single-step methods 45
6.2. Linear Multi-step procedure 47
Bibliography 49
3
CHAPTER 1
Time Systems
Many geodetic observation techniques use travel times of electromagnetic waves.
Therefore a precise denition of time is fundamental to geodetic observations.
Presently, two time systems are in use
atomic time,
dynamical time
Atomic time is the time scale which is derived from a set of terrestrial fundamental
atomic clocks. Dynamical time is the independent variable in the theory of General
Relativity. Due to the movement of the Earth around the galactic mass center the
atomic time diers periodically from the dynamic time. In order to give an example:
Atomic time (UTC or GPS-time) is used for the time tagging of the GPS-signal.
The equations of motion of a GPS satellite are formulated in dynamical time (DT).
Before atomic time was available civilian time systems were based on the Earths
rotation and were called universal (UT1) or sidereal (GMST,GAST). Universal or
sidereal time can be seen as an approximation of atomic time scale, which is strongly
coupled to the rotation of the Earth.
1.1. Atomic Time
The fundamental atomic time scale Temps Atomique International - TAI is
based on atomic clocks, operated by various national agencies and kept by the In-
ternational Earth Rotation Service - IERS and the Bureau International de Poids
et Mesures -BIPM).
Definition 1.1.1. TAI is a uniform time scale coinciding with Universal Time
(UT) at midnight January 1, 1958.
The fundamental interval unit of TAI is one SI second.
The SI day is dened as 86400 SI seconds and the Julian century is dened as 36525
SI days.
Definition 1.1.2. The Julian Date is the number of SI days and the fraction
of a SI day elapsed since 12, h UT on January 1, 4713 BC.
Since the origin of Julian Date is much to far in the past, UT has another
fundamental epoch to refer time dierences to:
Definition 1.1.3. The Julian Date of the standard epoch of UT is called
J2000.0. It is dened as the Julian Date at 1 January 2000 12:00 GMT.
Hence,
(1.1.1) J2000.0 = JD2451545.0
All time variables, denoted by T are measured in Julian centuries relative to the
epoch J2000.0.
5
6 1. TIME SYSTEMS
TAI is a continuous time scale, it does not maintain synchronization with the solar
day (UT) .Since the rotation rate of the Earth is slowing down the TAI will get
more and more ahead UT. This problem is solved by the denition of the Universal
Coordinated Time UTC.
Definition 1.1.4. UTC runs at the same rate as TAI but is incremented pe-
riodically by leap seconds.
Leap seconds are introduced by the IERS if necessary. The introduction of leap
seconds makes sure that the dierence between UTC and UT (more precisely: be-
tween UTC and UT1) is not larger than 0.9 s. The dierence DUT1 := UT1-UTC
is broadcasted by the IERS.
A third atomic time is the GPS time.
Definition 1.1.5. The GPS time (GPST) runs at the rate of the atomic clock
of the GPS Master Control station in Colorado Springs. GPST and UTC coincided
at o
h
January 6 1980.
Since GPST is not incremented there is a 19 seconds oset between TAI and GPST
(1.1.2) GPST = 19s +TAI
1.2. Dynamical Time
Definition 1.2.1. Dynamical Time is the independent variable in the equa-
tions of motion of bodies under gravitational forces according to the theory of
General Relativity.
Since the best approximation of an inertial system is centered at the barycenter of
the solar system, the dynamical time measured in this system is called Barycentric
Dynamical Time (Temps Dynamique Barycentrique - TDB).
An Earth based clock will show periodic variations of about 1.6 milliseconds with
respect to TDB due to the motion of the Earth in the gravitational eld of the sun.
Definition 1.2.2. Terrestrial Dynamical Time TDT (Temps Dynamique Ter-
restre) is the independent variable in the equation of motion of a body in the Earths
gravitational eld.
The relations between TDT and TDB are given by
(1.2.1) TDB = TDT + 0
s
.001658 sin(g + 0.0167 sin g)
where
(1.2.2) g = (357

.528 + 35999

.050T)

180
T is the time in Julian centuries TDT.
1.3. Sidereal and Universal Time
Prior to the operationality of atomic clocks, the Earths diurnal rotation was
used to measure time. Two dierent time systems were connected to the rotation
Earth:
sidereal time
universal time
1.3. SIDEREAL AND UNIVERSAL TIME 7
These two times are still used as an angle measure for the transformation between
celestial and terrestrial systems.
Definition 1.3.1. The angle between the observers local meridian and the true
vernal equinox corrected for precession and nutation is called apparent sidereal time
(AST).
If this angle is referred to the Greenwich mean astronomical meridian, it is called
Greenwich apparent sidereal time (GAST)
Besides the times AST and GAST, which refer to the true vernal equinox there are
corresponding times MST and GMST, which refer to the mean vernal equinox.(corrected
only for precession)
Definition 1.3.2. The dierence between GAST and GMST is called the equa-
tion of Equinox EqE
(1.3.1) Eq.E := GAST GMST = AST MST = cos( + )
where the nutations in longitude and obliquity , are given by (3.3.3) and
(3.3.4).
According to Keplers second law the Earth doesnt revolve the sun at a constant
angular velocity. For this reason a ctitious sun was invented which moves with
constant velocity.
Definition 1.3.3. The hour angle of the ctitious sun is called Universal Time
UT. The Time UT1 is the Universal Time corrected for polar motion.
The relationship between sidereal and universal time is given in terms of the
IAU(1967) system of constants by
(1.3.2)
GMST
0
= UT1 + 6
h
41
m
50

.548481 + 8640184//.812866T
u
+0

.093104T
2
u
6

.2 10
6
T
3
u
dGMST
dt
= 1.002737909350795 + 5.9006 10
11
T
u
5.9 10
15
T
2
u
GMST = GMST
0
+
dGMST
dt
UT1
With T
u
being the Julian date since J2000.0 in Julian centuries
(1.3.3) T
u
=
JulianUT1date 2451545.0
36525
CHAPTER 2
Conventional Reference Systems
2.1. International Earth Rotation Service (IERS)
Recent reference systems are maintained by international cooperation. The
International Association of Geodesy (IAG) has set up a service, the International
Earth Rotation Service (IERS), which is concerned with the maintenance of the
IERS Reference System. In dierent IERS Analysis Centers for geodetic space
techniques as VLBI, SLR LLR, GPS the parameters for the Earth rotation are
computed. In the Central Bureau these information are combined and in regular
updates the IERS Reference System is released.
The IERS Reference System is composed of
IERS Standards, a set of models and parameters, which are used by the
Analysis Centers.
the IERS Celestial Reference Frame (ICRF) and
the IERS Terrestrial Reference Frame (ITRF)
The ICRF is realized by a catalogue of compact extragalactical radio sources. The
ITRF is realized by a set of terrestrial station coordinates and velocities.
2.2. Celestial Reference System
Definition 2.2.1. The small motions of the Earths rotation axis can be de-
scribed as the sum of two components
(1) astronomical nutation
(2) polar motion
The direction of the axis which is computed from the theory of nutation and
precession is called Celestial Ephemeris Pole (CEP)
The origin of the ICRS is the barycenter of the solar system.
The axes of the ICRS are dened as
the CEP,
the equinox
and a third axes completing the former two axis to a Cartesian coordinate
system
at the epoch J2000.0
The ICRF is a realization of the IERS consisting of catalogue of astronomical co-
ordinates of about 200 extragalactical radio sources at the epoch J2000.0.
By adopting coordinates of quasars, implicitly a coordinate system is conventionally
dened. The conventionally dened coordinate system diers in the orientation of
its axis by about 0.0001 arc-seconds from the ICRS.
The transformation from the ICRF to a system with its third axis to CEP is given
by the theory of nutation and precession.
2.3. Terrestrial Reference Systems
The CEP moves with respect to the Earths surface. In order to have a coor-
dinate system , which is xed with respect to the Earth the ITRS is adopted.
9
10 2. CONVENTIONAL REFERENCE SYSTEMS
Figure 2.2.1. Equinox
Definition 2.3.1. The mean direction of the Earth rotation axis determined
by the ve International Latitude Service stations in the period 1900.0 to 1906.0 is
dened as the Conventional Inertial Pole (CIO) at the epoch 1903.0
Definition 2.3.2. The ITRS is dened with its origin at the Earths geocenter.
The axes of the ITRS are oriented in the following way:
the Z axis is oriented to the CIO,
the X-axis is oriented towards the 1903.0 meridian of Greenwich
and the Y axis completes the former two axes to a Cartesian coordinate
system.
The ITRS is realized by the ITRF, a catalogue of Cartesian coordinates and
velocities of globally distributed tracking stations.
The adopted coordinates of these stations implicitly dene a new coordinate system,
which diers from the ITRS by about 10mm in position and several mm/year in
velocity.
The transformation between the ITRF and the ICRF is given by the pole coordi-
nates x
P
, y
P
and the nutation and precession parameters d, d.
2.4. WGS84
Besides the ITRF several other terrestrial reference systems are in use. Prob-
ably, the most important is the WGS84. The WGS84 is maintained by the US
Department of Defense (DoD) and is the reference system of the GPS system. It
is an implicitly dened system. It is dened by adopting Cartesian coordinates of
the ten DoD GPS Monitoring Stations derived from Doppler observations on these
sites. This results in an accuracy of the WGS84 System of about 1-2 m. In order
to align the WGS84 with the more accurate ITRF the DoD has recomputed the
coordinates of the ten monitoring stations using GPS observations at these sites
2.4. WGS84 11
Figure 2.3.1. Relation between ICRS and ITRS
Figure 2.3.2. ITRF97 Network
and at a subset of IGS tracking stations whose ITRF coordinates were held xed.
This rened WGS84 System is called WGS84 (G730).
The WGS84 system is realized by the ephemerides of the GPS satellites. In order to
compute the orbits of these satellites some additional constants have to be adopted.
12 2. CONVENTIONAL REFERENCE SYSTEMS
An ellipsoidal coordinate system is attached to the WGS84 by locating an ellipsoid
at the origin of the WGS84 system and letting the rotation axis coincide with the
Z-axis of the WGS84. This means the datum parameters of the WGS84 are
Table 1. Datum parameters of WGS84
Parameter Symbol numerical value
semi-major axis a 6378137 m
reciprocal attening 1/f 298.257223563
angular velocity 7.292115 10
5
s
1
geocentric
gravitational constant GM 398600.5 km
3
s
2
second zonal harmonic

C
2,0
4884.16685 10
6
2.5. Ellipsoidal Reference Systems
For many purposes ellipsoidal coordinate systems are more convenient than
Cartesian systems. Ellipsoidal systems can be distinguished between global systems
which approximate the Earth as a whole and a local ellipsoidal system, which
approximates the Earths surface in a certain region.
2.5.1. The GRS80 Reference System. The most important global ellip-
soidal System is the GRS System. It is dened as an ellipsoid centered at the
origin of the ITRS and having its axes coinciding with the axes of the ITRS. The
additional datum parameters of the GRS80 Reference System are
Table 2. Datum parameters of GRS80
Parameter Symbol numerical value
semi-major axis a 6378137 m
reciprocal attening 1/f 298.257222100827
angular velocity 7.292115 10
5
s
1
geocentric
gravitational constant GM 398600.5 km
3
s
2
dynamical form factor J
2
108263 10
8
The GRS80 Reference system is a global ellipsoidal system. It approximates
the Earth as a whole. Besides global ellipsoidal systems a number of local ellipsoidal
systems are in use. They approximate the Earths surface only in their region of
validity.
2.5.2. Local Ellipsoidal Systems.
2.5.2.1. The Rauenberg Datum. The Rauenberg datum is the ocial reference
system for the western part of Germany. It is an ellipsoidal System which is based
on the Bessel 1841 ellipsoid. This ellipsoid has the dimensions
(2.5.1) a = 6377397.155 m, 1/f = 299.15281285
The position and orientation datum parameters are not given as the position of
the origin and the orientation of the axes but in an equivalent way. First an initial
point in the center of the region of validity of the reference system has to be xed.
For this initial point the following quantities are assigned
2.5. ELLIPSOIDAL REFERENCE SYSTEMS 13
the ellipsoidal coordinates L, B are set identical to the astronomic coor-
dinates , ,
the geodetic azimuth A to a specic target is set identical to the astro-
nomical azimuth a
a specic value N is adopted for the separation between the geoid and the
ellipsoid in the initial point.
For the Rauenberg datum the Helmert Tower of the GeoForschungsZentrum in
Postsdam was used as initial point. Its ellipsoidal coordinates were set to
(2.5.2) B = 52

22

53

.9540 N, L = 11

34

26

.483 E
The coordinate system was oriented by setting the geodetic azimuth of the line
Rauenberg - Marienkirche Berlin to its astronomical azimuth.
The reference frame of the Rauenberg datum are the points of the German Haupt-
dreiecksnetz (DHDN) whose Gau - Kr uger coordinates based on a 3

zone-width
are given.
Figure 2.5.1. Deutsches Hauptdreiecksnetz
2.5.2.2. The 42/83 Datum. The system 42/83 used to be the ocial reference
system in the eastern part of Germany. It is based on the Krassovsky 1940 ellipsoid
with the following dimensions
(2.5.3) a = 6378245 m, 1/f = 298.3
14 2. CONVENTIONAL REFERENCE SYSTEMS
Its initial point is the Central Astronomical Observatory Pulkovo (close to St. Pe-
tersburg) and its reference frame are the points of
the Einheitliches Astronomisch-Geodatisches Netz(EAGN) and
the Staatliches Trigonometrisches Netz 1st Order (STN1.O)
whose Gau - Kr uger coordinates based on a 6

zone-width are given.

CHAPTER 3
Earth Rotation
The rotation axis of the Earth is not xed in inertial space neither it is with
respect to the Earths body. The gravitative forces of the Sun and Moon acting
on the equatorial bulge of the Earth are changing the orientation of the rotation
axis in inertial space. These changes are called precession and nutation and can be
predicted with a very high accuracy.
Additionally, there is a small movement of the Earths rotation axis with respect
to its crust, which is called polar motion. Both nutation and polar motion are the
superposition of Earths response to external forces and free nutation of the Earths
Nutation is primarily the forced response of the Earth and can be predicted by
geophysical and orbital models. The polar motion represents the forced and the
free response of the Earth to external forces in almost equal parts. Again the forced
part can be predicted but the free part can only be determined by Space Geodesy
methods.
3.1. Motion in Celestial System
Moon and Sun and the planets exert gravitational forces on the equatorial
bulge. Since the rotating Earth behaves like a gyro, it reacts to this forces by a
clockwise movement of its rotation axis. This movement consist of two constituents
precession and
nutation
Theorem 3.1.1. Luni-solar precession is the circular motion of the celestial
pole with a period of 25, 800 years and an amplitude equal to the obliquity of the
ecliptic of 23

.5. The precession causes a westerly movement of the equinox of about

50

.3 per year.
Planetary precession consist of a 0

.5 per year rotation of the ecliptic resulting

in a easterly motion of the equinox by about 12

.5 per century and an decrease of

the obliquity of the ecliptic by about 47

per century.
Definition 3.1.2. The combined eect of luni-solar and planetary precession
is called general precession or simply precession
Definition 3.1.3. The short periodic motion of the pole superimposed on the
precession with oscillations of 1 day to 18.6 years (the main period) and a maximum
amplitude of 9

.2 is called nutation.
3.2. Precession transformation
The transformation of stellar coordinates from the mean equator and equinox
at epoch t
i
to the mean equator and equinox at another epoch t
j
is performed by
the means of the following rotation matrix
(3.2.1) P = R
3
(z
A
)R
2
(
A
)R
3
(
A
)
15
16 3. EARTH ROTATION
Figure 3.1.1. Precession cone
Figure 3.1.2. Nutation
The precession angles , dened by the 1976 IAU conventions, are given by
(3.2.2)

A
= (2306

.2181 + 1

.39656T
u
0

.000139T
2
u
)t
+(0

.30188 0

.000344T
u
)t
2
+ 0

.017998t
3
(3.2.3)
z
A
= (2306

.2181 + 1

.39656T
u
0

.000139T
2
u
)t
+(1

.09468 0

.000066T
u
)t
2
+ 0

.018203t
3
(3.2.4)

A
= (2004

.3109 0

.85330T
u
0

.000217T
2
u
)t
(0

.42665 0

.000217T
u
)t
2
0

.041833t
3
where
(3.2.5) T
u
:= (JD 2451545.0)/36525
and t is the interval between t
j
and t
i
in Julian centuries.
3.3. Nutation transformation
The transformation of stellar coordinates from the mean to the true equator
and equinox at an epoch is given by
(3.3.1) N = R
1
( )R
3
()R
1
()
3.3. NUTATION TRANSFORMATION 17
The nutation time series according to the 1980 IAU conventions are
(3.3.2)
= (84381

.448 46

.8150T
u
+ 0

.00059T
2
u
+ 0

.001813T
3
u
)
+(46

.8150 0

.00177T
u
+ 0

.005439T
2
u
)t
+(0

.00059 + 0

.005439T
u
)t
2
+ 0

.00181t
3
The nutation parameters and can be represented by series expansions
(3.3.3) =
N

j=1
_
(A
0j
+A
1j
T) sin
_
5

i=1
k
ji

i
(T)
_
)
_
(3.3.4) =
N

j=1
_
(B
0j
+B
1j
T) sin
_
5

i=1
k
ji

i
(T)
_
)
_
The coecients are arguments of the motion of Sun and Moon:
(1) mean anomaly of the Moon
(3.3.5)
1
= 485866

.733 + (1325
r
+ 715922

.633)T + 31, //.310T

2
+ 0

.064T
3
(2) mean anomaly of the Sun
(3.3.6)
2
= 1287009

.804 + (99
r
+ 1292581

.224)T 0

.577T
2
0

.012T
3
(3) mean argument of latitude of the Moon
(3.3.7)
3
= 335778

.877 + (1342
r
+ 2995263

.137)T 13

.257T
2
+ 0

.011T
3
(4) mean elongation of the Moon from the Sun
(3.3.8)
4
= 10072261

.307 + (1236
r
+ 1105601

.328)T 6

.891T
2
+ 0

.019T
3
(5) mean longitude of the ascending lunar node
(3.3.9)
5
= 450160

.280 (5
r
+ 482890

.539)T + 7

.455T
2
+ 0

.008T
3
Here 1
r
means one revolution , i.e. 1
r
= 360

= 1296000

. The coecients
A
ij
, B
ij
, k
ij
are given by the standard 1980 IAU series and can be found in [4].
3.3.1. Motion in the Terrestrial System. Besides the movement of the
Earths rotation axis in space there is an additional variation of the rotation axis
relative to the Earths crust. This motion is primarily due to the elastic properties
of the Earth and due to the exchange of angular momentums between the solid
Earth, the oceans and the atmosphere.
Definition 3.3.1. Polar motion is the rotation of the true celestial pole as
dened by the precession and nutation models with respect do the z-axis of a
conventionally chosen terrestrial reference system.
Polar motion consists of a free and a forced oscillation. The free oscillation is
counterclockwise with a period of 430 days (Chandler period) and an amplitude of
3-6 m.
The forced component again consists of two parts. One part is excited by the
tidal forces and therefore has a diurnal period, with an amplitude of one order of
magnitude smaller then the free oscillation. The second part has an annual period
since it is excited by the annual changes in the atmosphere. Its amplitude is about
as large as the free oscillation.
Polar motion cannot be predicted by models, it has to be observed by space tech-
niques. The accuracy of those observation has achieved a very high level, accounting
for 0.2 - 0.5 milliarcseconds which is equivalent to 6 - 15 mm at the Earths surface.
Polar motion values can be downloaded from the International Earth Rotation Ser-
vice (IERS) as tables of daily values of pole coordinates.
18 3. EARTH ROTATION
3.3.1.1. Earth Orientation Transformation. The transformation from the ce-
lestial to the terrestrial system includes the Earths rotation and the polar motion.
Consequently it can by represented as the product of three rotation matrices:
(3.3.10) S = R
2
(x
p
)R
1
(y
p
)R
3
(GAST)
The Earth rotation is the rotation around the instantaneous rotation axis with the
rotation angle being the dierence between the true vernal equinox of the date to
the meridian of the 1903.0 Greenwich zero longitude.
(3.3.11) R
3
() =
_
_
cos sin 0
sin cos 0
0 0 1
_
_
, = GAST
where GAST is given by
(3.3.12) GAST = GMST
0
+
d(GMST)
dt
(UTC) (UTC UT1) +Eq.E
with Eq.E being again the equation of equinox. Here the dierence UTC UT1
has to be interpolated from IERS tables.
The polar motion rotation is the transformation between the instantaneous pole to
the pole given by nutation and precession theories.
Definition 3.3.2. Polar motion rotation is dened by the left-handed pair of
angles (x
p
, y
p
) . The rst angle is the angle between the mean direction of the pole
during the period 1900.0 - 1906.0 (the IERS Reference Pole (IRF)) and the true
rotation axis. It is dened positive in the direction of the x-axis of the terrestrial
system. The second angle is positive in the direction of the 270

meridian.
Since both angles are small the rotation can be approximated by
(3.3.13) R
2
(x
p
)R
1
(y
p
) =
_
_
1 0 x
p
0 1 y
p
x
p
y
p
1
_
_
where the angles are interpolated from the IERS tables.
Figure 3.3.1. Polar motion
CHAPTER 4
Signal Propagation
Signals emitted from the satellite travel through ionosphere and troposphere.
Due to the interaction of the electromagnetic signal with ionosphere and tropo-
sphere
the signal strength,
the signal direction and
the velocity of wave propagation
will be altered. In this section some fundamentals of signal propagation will be given
and the characteristics of the ionospheric and tropospheric signal disturbance will
be discussed. Detailed correction formulas for particular observation techniques will
be discussed in connection with the description of the corresponding observation
technique.
4.0.2. Fundamentals of Wave Propagation.
4.0.2.1. Basic Relations and Denitions. In a reasonable degree of accuracy
the scalar wave equation
(4.0.14)
_

1
c(x)

2
t
_
U(t, x) = 0
can be used as a mathematical model for wave propagation. In(4.0.14) U stands
for each component of the electric and magnetic eld in free space and c(x) is the
position dependent wave speed. As the rst approximation we solve (4.0.14) for a
constant propagation speed c(x) = c.
(4.0.15)
_

1
c

2
t
_
U(t, x) = 0
Equation (4.0.15) describes the eld generated without an external source only
generated from the initial conditions at t = 0. If a delta function point source at
the origin and at the time zero is considered as the cause for the electromagnetic
eld G
0
, the eld solves
(4.0.16)
_

1
c

2
t
_
G
0
(t, x) = (t)(x)
Equation (4.0.16) has the solution
(4.0.17) G
0
(t, x) =
(t
|x|
c
)
4[x[
c.f. [1]. This means at a specic epoch t

the delta pulse reaches all points with

the distance c t

from the origin. The solution (4.0.16) is a spherical wave with

the travel velocity c.
If the delta source is replaced by a periodic source at the origin the corresponding
wave equation is
(4.0.18)
_

1
c

2
t
_
U(t, x) = e
t
(x)
19
20 4. SIGNAL PROPAGATION
having the solution
(4.0.19) U(t, x) = G
0
(t, x) e
t
(x) =
e
(t
|x|
c
)
4[x[
This again is a spherical wave with the velocity c. For a xed location x the eld
U has a periodic behavior with the period
(4.0.20) P =
2

The reciprocal value of the period is called frequency

(4.0.21) f =
1
P
it characterizes the number of repetitions of the same state per second at a xed
location. Frequency f and circular frequency are related to each other by
(4.0.22) = 2f
The distance of two points where for a xed epoch t the wave has the same state is
called the wave-length . It equals the distance the wave travels within one period
P. Consequently
(4.0.23) = c P =
2c

=
c
f
The quantity
(4.0.24) k :=
2

is the so called wave number. The constant c is the vacuum propagation velocity
and its numerical value is adopted by scientic communities. The value currently
in use in satellite geodesy is
(4.0.25) c = 2.99792458 10
8
ms
1
.
If the wave travels through a medium dierent form the vacuum its velocity changes
from c to a dierent value v. the ratio
(4.0.26) n :=
c
v
is called index of refraction. Besides the index of refraction in some cases also the
refractivity
(4.0.27) N = (n 1) 10
6
is used for a characterization of the mediums electromagnetic properties.
4.0.2.2. Dispersion, Phase Velocity and Group Velocity. The velocity v

a wave
travels through a medium is called phase velocity. In a dispersive medium the
phase velocity depends on the frequency. If a superposition of dierent frequencies
travels through a dispersive medium besides the phase velocity a second velocity is
produced, the so-called group velocity. In order to explain what a group velocity
is, a simple example of two waves with dierent frequencies travelling through a
dispersive medium is considered.
s(t, x) = e
[(+)t(k+k)|x|]
+e
[()t(kk)|x|]
= e
[()t(k+k)|x|]
e
t
+e
[()t(kk)|x|]
e
t
= e
[()t(k)|x|]
e
[tk|x|]
+e
[()t(k)|x|]
e
[tk|x|]
(4.0.28)
= 2e
[()t(k)|x|]
cos(t k[x[)
The superposition of waves with dierent frequencies travelling with dier-
ent velocities k k results in a wave of frequency travelling with the velocity
4.1. STRUCTURE OF THE ATMOSPHERE 21
/k which is modulated by a wave of frequency , which travels with the velocity
/k. The velocity of this articial modulating wave is called group velocity.
For an animation click here
From the simple example can be concluded that
(4.0.29) v
g
=
d
dk
holds. Let us now assume that the phase velocity v
p
depends on the wave-length
v
p
= v
p
()
with these two relationships a connection between phase velocity v
p
and group
velocity can be derived
(4.0.30)
v
g
=
d
dk
=
2d(
v
p
()

)
d(
2

)
=
1

2
(
v
p

v
p
)d

2
d
= v
p

v
p

A medium is called dispersive if the phase velocity depends on the frequency. The
ionosphere is a dispersive medium but the troposphere is not.
From relation (4.0.30) a corresponding relation between the refraction indices of
phase and group can be derived easily. Obviously,
n
g
=
v
g
c
=
v
p
c

dn
p
d
= n
p

v
p
f
dn
p
d
holds. Since
d = d(
v
p
f
) =
v
p
f
2
df
is valid, the nal result is
(4.0.31) n
g
= n
p
+f
dn
p
df
4.1. Structure of the atmosphere
With respect to signal propagation the atmosphere is usually divided into tro-
posphere and ionosphere
The troposphere is the lower part of the atmosphere which extends from
the surface of the Earth to about 40 km above. The signal propagation
in the troposphere depends mainly on the water vapor content and the
temperature.
The ionosphere is the upper part of the atmosphere extending approxi-
mately between 70 km and 1000 km above the Earth. The signal propaga-
tion in the ionosphere is mainly inuenced by the density of free electrons.
The troposphere is the gaseous part of the atmosphere charged particles are practi-
cally absent, which means that the troposphere can be considered electrically neu-
tral. In the troposphere the refraction index is slightly larger than 1. It decreases
22 4. SIGNAL PROPAGATION
with increasing height. The state of the ionosphere is described by its electron den-
sity n
e
in units of [number of electron m
3
]. The refraction index in the ionosphere
depends on the electron density n
e
in the following way [2]
(4.1.1) n = 1 C
n
e
f
2
, C = 40.3 m
3
s
2
Formula (4.1.1) shows that the time delay of signal propagation is indirectly propor-
tional to the square of frequency and therefore decreases for increasing frequencies.
4.1.1. Signal propagation through ionosphere and troposphere.
4.1.1.1. Ionospheric refraction. The inuence of the ionosphere on signal prop-
agation is characterizes by dispersion. The refraction index of the phase delay is
given by
n
p
= 1 40.3
n
e
f
2
using (4.0.31) the corresponding refraction index of the phase delay is obtained by
(4.1.2)
n
g
= n
p
+f
dn
p
df
= 1 40.3
n
e
f
2
+ 2f
40.3n
e
f
3
= 1 + 40.3
n
e
f
2
A comparison of (4.1.1) and (4.1.2) shows that the inuence of the ionosphere on
the group velocity and the phase velocity has the same magnitude but the opposite
sign. The ionospheric delay induce range errors versus the frequency is given in the
following Table.
Table 1. Inuence of ionosphere onto ranges
f 400 MHz 1600 MHz 2000 MHz 8000 MHz
average 50 m 3 m 2 m 0.12 m
90% smaller than 250 m 15 m 10 m 0.6 m
maximum 500 m 30 m 20 m 1.2 m
4.1.1.2. Tropospheric refraction. Refractive eects in the troposphere are inde-
pendent of the frequency. The refractivity at the base of the troposphere is given
by
N
T
= C
1
P
T
+C
4
e
T
2
= N
d
+N
w
where
P is the air pressure in [HPa]
e is the partial pressure of water vapor in [HPa]
and T is the temperature in [K]
Recommended numerical values are
C
1
= 77.6, C
4
= 3.73 10
5
Obviously, the inuence of the troposphere can be divided into
N
d
- the dry part and
N
w
- the wet part.
The resulting range error can be computed by
(4.1.3) S = S
d
+ S
w
= 10
6
_
H
d
R
N
d
dS + 10
6
_
H
w
R
N
w
dS
Where the integral is evaluated along the signal path from the observer R upto the
eective altitudes H
d
, H
w
of the dry and wet term.
4.1. STRUCTURE OF THE ATMOSPHERE 23
Since neither the air pressure P nor the water vapor pressure e can be measured
along the signal path these missing data have to be replaced by an empirical model.
The Hopeld model [3] is widely used
(4.1.4) N
d
(h) = N
d0
_
H
d
h
H
_
4
where
H
d
= 40.136 + 148.72(T 273.16)
and
(4.1.5) N
w
(h) = N
w0
_
H
w
h
H
_
4
, H
w
= 11000 m
with N
d0
, N
w0
being the dry and the wet refractivity at height-zero. Introducing
(4.1.4) and (4.1.5) into the integral (4.1.3) and assuming the signal path to be a
straight line yields
S
T
=
K
d
_
sin(E
2
+ 6.25)
+
K
w
_
sin(E
2
+ 2.25)
With
K
d
= 155.2 10
7 P
T
H
d
K
w
= 155.2 10
7 4810e
T
2
H
w
an E being the elevation angle of the incoming signal measured in degrees.
The inuence of tropospheric refraction onto measured ranges is indicated in the
following Table.
Table 2. Inuence of troposphere onto ranges
E 90

20

15

10

S
d
2.31 m 6.71 m 8.81 m 12.90 m 23.61 m
S
w
0.20 m 0.58 m 0.77 m 1.14 m 2.21 m
S
T
2.51 m 7.29 m 9.58 m 14.04 m 25.82 m
CHAPTER 5
Theory of Satellite Motion
5.1. Equation of Motion
The motion of an articial satellite is determined by the gravitational eld of
the Earth, Moon, Sun and the planets. In this form the satellite motion is the
solution of the multi-body problem in celestial mechanics. Since there is no closed
solution of this problem some simplication will be made here
(1) the gravitational eld of the Earth, Sun, Moon and the planets are con-
sidered point-mass elds
(2) the gravitational eld of the satellite itself is neglected
(3) the gravitational elds are considered time-invariant, i.e. all tidal potential
are neglected
Under these assumptions the equation of motion for the i-th body is:
(5.1.1) m
i
x
i
= k
with m
i
being the mass of the i-th body and k being the gravitational force of all
other bodies acting on the i-th body
(5.1.2) k =
n

j=1
k
ij
= G
n

j=1
m
i
m
j
|x
i
x
j
|
3
(x
i
x
j
) j ,= i
The equation
(5.1.3) m
i
x
i
= G
n

j=1
m
i
m
j
|x
i
x
j
|
3
(x
i
x
j
) j ,= i
describes the motion of the i-th body in an inertial fame. If the motion is considered
relative to the motion of the n-th body, (5.1.3) changes into
(5.1.4)
x
i
x
n
= G

n
j=1
m
j
(x
i
x
j
)
x
i
x
j

3
+G

n1
j=1
m
j
(x
n
x
j
)
x
n
x
j

3
= G(m
n
+m
i
)
x
i
x
n
x
i
x
n

3
+

n1
j=1
m
j
_
x
i
x
j
x
i
x
j

3
+
x
j
x
n
x
j
x
n

3
_
j ,= i
Introducing the notations
(5.1.5) X
i
= x
i
x
n
, M
i
= m
n
+m
i
yields
(5.1.6)

X
i
= GM
i
X
i
|X
i
|
+G
n1

j=1
m
j
_
x
i
x
j
|x
i
x
j
|
3
+
X
j
|X
j
|
3
_
j ,= i
The body i can be identied with the satellite and the body n with the Earth.
Therefore, the rst term in (5.1.6) describes the gravitational force of the Earth
and the second term the gravitational forces of the other bodies acting on the
25
26 5. THEORY OF SATELLITE MOTION
satellite. Since the distance of the other bodies from the satellite is much larger
than the distance of the satellite from the Earth, the following relations are true
m
j
x
j
x
i

2

M
i
X
i

2
,
m
j
X
j

2

M
i
X
i

2
This leads to the following simplication of (5.1.6)
(5.1.7)

X
i
= Gm
n
X
i
|X
i
|
3
This equation is called the Keplerian equation of motion. It describes the motion
of a satellite in the absence of any other attracting body and under the assumption
the Earth being a homogeneous spherical body with the mass m
n
.
The real satellite orbit diers from the solution of (5.1.7) due to
the deviation of the Earth from a homogeneous spherical body
the gravitational inuence of Moon, Sun and the planets
the radiation pressure of the sun
atmospheric friction
tides
polar motion
5.2. Integration of Keplerian equation of motion
The dierential equation (5.1.7) is a nonlinear second order dierential equa-
tion. The solution of (5.1.7) unambiguous, if 6 integration constants are xed.
These constants could be the initial position and the initial velocity or any other
six functions, which are constant along the satellite orbit.
Since the equation (5.1.7) is nonlinear it can not be solved straight-forwardly but
the so-called method of rst integrals has to be applied. Method of rst integrals
means to identify six quantities C
i
which are constant along the orbit.
(5.2.1) C
k
= C
k
_
X
i
,

X
i
, t
_
= const, k = 1, . . . , 6
The constants C
k
are then used as integration constants. Having identied the rst
integrals C
k
a solution of (5.2.1) for X
i
,

X
i
delivers the orbit.
The methods of the rst integrals will be now applied in detail to equation (5.1.7),
which for simplicity is rewritten as
(5.2.2)

X = GM
X
|X|
3
Step 1: Equation (5.2.2) is multiplied vectorial with X
(5.2.3) X

X =
GM
|X|
3
(X X) = 0
Since
(5.2.4) 0 = X

X =
d
dt
(X

X)
the relation
(5.2.5) C = X

X
immediately follows. Equation (5.2.5) represents the rst two integrals of (5.2.2),
since it proves that for a Keplerian motion the orbital spin vector
(5.2.6) I := X m

X = const
is constant. This property corresponds to Keplers second law:
The planetary motion is in a plane.
5.2. INTEGRATION OF KEPLERIAN EQUATION OF MOTION 27
The orientation of the orbital plane in space is given by the two angles
i -the inclination
-the right ascension of ascending node
The denition of these two angles can be read from the following gure. Hence,
Figure 5.2.1. orbital geometry
and i are the rst two integration constants. Since
(5.2.7) C = |C| (sin i sin e
1
sin i cos e
2
+ cos ie
3
)
the third integration constant is |C|.
Step 2: Vectorial multiplication of (5.2.2) with C
(5.2.8)

X C =
GM
|X|
3
_
X (X

X)
_
or equivalently
(5.2.9)
d
dt
_

X C
_
=
GM
|X|
3
_
X(X
T

X)

X(X
T
X)
_
Since
d
dt
|X| =
X
T

X
|X|
from (5.2.9) follows
(5.2.10)
d
dt
_

X C
_
= GM
d
dt
|X| X

X|X|
|X|
2
= GM
d
dt
_
X
|X|
_
28 5. THEORY OF SATELLITE MOTION
Integration on both sides yields
(5.2.11)

X C = GM
X
|X|
+P
with P being the integration constant. Since P has to be orthogonal to C it has
only two degree of freedom, which are the next two integration constants of the
Kepler problem. These two degrees of P are
- its length |P| and
- the angle between P and the nodal line.
The angle is called argument of perigee and the straight line passing through P
is called apsidal line.
Step 3: Scalar multiplication of

X C and X
(5.2.12) X
T
(

X C) = X
T
_
GM
X
|X|
+P
_
Since the dot-product X
T
(

X C) is invariant against cyclic permutation of it
arguments, the relation
(5.2.13) C
T
(X

X) = GM
X
T
X
|X|
+X
T
P
follows. Which after slight manipulation becomes
(5.2.14) |C|
2
= GM |X|
_
1 +
|P|
GM
_
X
|X|
_
T
P
|P|
_
If the angle between X and P is denoted by the nal relation
(5.2.15) |X| =
C
2
GM
1 +
P
GM
cos
is obtained, which gets its nal form by the introduction of the abbreviations p and
e:
(5.2.16) |X| =
p
1+e cos
, p =
C
2
GM
, e =
P
GM
Equation (5.2.16) is the equation of a cone-section in polar form.
This means that (5.2.16) reects Keplers rst law
The planetary motion is a cone section.
The kind of the cone-section is indicated by the numerical value of e, the so called
numerical eccentricity.
e = 0 circle
0 e < 1 ellipse
e = 1 parabola
e > 1 hyperbola
All geodetic relevant satellites have very small numerical eccentricity.
So far, the shape of the satellite motion and its location in space has been deter-
mined. The last step will be the establishment of the timely evolution of the motion
of the satellite along the cone-section; i.e. the determination of the angle as a
function of time: = (t).
Step 4: As an auxiliary step, we represent the orbit in polar coordinates with
respect to an orbital-plane attached coordinate system. Its origin coincides with
the focal point, its axis points to the Earth-closest point, the so-called perigee, the
5.2. INTEGRATION OF KEPLERIAN EQUATION OF MOTION 29
Figure 5.2.2. Cartesian orbital system
-axis is in the orbital plane and orthogonal to the -axis and the -axis is normal
to the orbital plane.
With respect to this coordinate system the satellite position is
X = |X| (cos d
1
+ sin d
2
) =: |X|d

X =
dX
dt
d +|X|
dd
dt
Hence
C = X

X = |X|d
_
dX
dt
d +|X|

d
_
= |X|
2
d

d
= |X|
2
(cos d
1
+ sin d
2
) (sin d
1
+ cos d
2
)
= |X|
2

_
cos
2
+ sin
2

_
d
1
d
2
= |X|
2
d
3
Computing the norm on both sides yields
(5.2.17) |C| = |X|
2
Equation (5.2.17) is the so-called area theorem or Keplers second law
The line from the Earth to the satellite covers identical areas in
identical times.
Equation(5.2.17) is also a dierential equation for = (t). It is solved by separa-
tion of the variable
|C| dt = |X|
2
d
leading to
|C| dt =
p
2
(1 +e cos )
2
d
30 5. THEORY OF SATELLITE MOTION
Integration on both sides gives
(5.2.18) |C| (t t
0
) =
_

0
p
2
(1 +e cos )
2
d
solving (5.2.18) for the upper limit gives the solution = (t). Unfortunately, a
closed solution in this form is impossible. The integral in (5.2.18) is a so-called
elliptic integral, which does not have a solution in closed form.
In order to solve (5.2.18) an auxiliary quantity, the so-called eccentric anomaly is
introduced. The denition of the eccentric anomaly can be read from Fig. (5.2.3).
Figure 5.2.3. true and eccentric anomaly
Obviously,
(5.2.19) |X| cos = a cos E ae
holds. Therefore
|X| =
p
1 +e cos
=
a cos E ae
cos
,
which leads to
(5.2.20)
p cos = a(cos E e) +ae cos (cos E e)
cos [p ae(cos E e)] = a(cos E e)
cos =
a(cos Ee)
pae(cos Ee)
=
cos Ee
p
a
e(cos Ee)
Using the relationship
p
a
= (1 e
2
) yields the nal result.
(5.2.21) cos =
cos Ee
1e cos E
cos E =
cos +e
1+e cos
5.2. INTEGRATION OF KEPLERIAN EQUATION OF MOTION 31
(5.2.22) sin =

1e
2
sin E
1e cos E
sin E =

1e
2
sin
1+e cos
Equations (5.2.21) and (5.2.22) can be used to substitute cos in the integral
(5.2.18) by a function of E. For a complete substitution also the replacement
of the dierential d by dE is necessary.
(5.2.23)
d
dE
sin =
_
d
d
sin
_
d
dE
= cos
d
dE
On the other hand
(5.2.24)
d
dE
sin =
d
dE
_

1e
2
sin E
1e cos E
_
=

1e
2
cos E
1e cos E

1e
2
sin Ee sin E
(1e cos E)
2
=

1e
2
cos E

1e
2
e cos
2
E

1e
2
e sin
2
E
(1e cos E)
2
=

1e
2
cos E

1e
2
e
(1e cos E)
2
=

1e
2
(cos Ee)
(1e cos E)
2
=

1e
2
cos
(1e cos E)
A comparison of (5.2.23) and (5.2.24) yields
(5.2.25) d =

1 e
2
1 e cos E
dE
Inserting (5.2.21) and (5.2.25) into (5.2.18) leads to
(5.2.26)
|C| (t t
0
) =
_
E
E
0
p
2
(1+e
cos Ee
1e cos E
)
2

1e
2
1e cos E
dE
=
_
E
E
0
p
2

1e
2
(1e cos E)
(1+e cos E+e cos Ee
2
)
2
dE
=
_
E
E
0
p
2
(1e
2
)
2

1 e
2
(1 e cos E)dE
= a
2

1 e
2
_
E
E
0
(1 e cos E)dE
In this way the elliptic integral (5.2.18) was replaced by the simple trigonometric
integral (5.2.26). The evaluation of (5.2.26) is very simple
(5.2.27) |C| (t t
0
) = a
2
_
1 e
2
(E e sin E)

E
E
0
If we now recall the denition of p
(5.2.28) p =
|C|
2
GM
we obtain
(E e sin E)

E
E
0
=
C
a
2

1e
2
(t t
0
) =
_
C
2
a
3
(1e
2
)
(t t
0
)
=
_
C
2
a
3
p
(t t
0
) =
_
GM
a
3
(t t
0
)
Without any loss of generality the time t
0
can be chosen as that instant of time the
satellite passes the perigee. For this choice E
0
= E(t
0
) = 0 holds, leading to
(5.2.29) E e sin E =
_
GM
a
3
(t t
0
)
32 5. THEORY OF SATELLITE MOTION
Obviously
E(t) = E
_
t + 2k
_
a
3
GM
_
k = 0, 1, 2
holds. This means after
(5.2.30) T = 2
_
a
3
GM
the satellite position repeats itself. T is called orbital period and relation (5.2.30)
reects Keplers third law
The squares of the orbital periods are proportional to the third
power of the semi-major axes.
The mean angular velocity of the satellite is given by
(5.2.31) n =
2
T
=
_
GM
a
3
Equation (5.2.31) motivates the denition of a third angular quantity- the mean
anomaly M,
(5.2.32) M = n(t t
0
)
The mean anomaly M is the true anomaly of a ctive satellite orbiting the Earth
on a circular orbit with the same orbital period as the actual satellite. Combining
(5.2.32) and (5.2.29) gives
(5.2.33) E e sin E = M
Equation (5.2.33) is the so-called Kepler-equation. It connects the given mean
anomaly M with the unknown eccentric anomaly E.
Since (5.2.33) is a nonlinear equation, it has to be solved iteratively. Due to the fact
that the numerical eccentricity e is very small, the iteration converges very fast.
Now all six integration constant of the Kepler-problem are found. They are the
so-called Keplerian elements
a, e, i, , , t
0
The rst two elements a, e describe shape and size of the orbital ellipse. The next
two elements , i describe the orientation of the orbital plane in space. The element
denes the orientation of the orbital ellipse in the orbital plane and t
0
is the time
the satellite passes the perigee.
5.2.1. Computation of satellite position from Keplerian elements.
First for a given time t the corresponding mean anomaly M(t) is computed
(5.2.34) M(t) = n(t t
0
), n =
_
GM
a
3
Using Kepler-equation the mean anomaly is converted into the eccentric anomaly
(5.2.35)
E
0
(t) = M(t)
E
n+1
= M(t) +e sin E
n
(t), n = 0, 1, 2, . . .
The sine and the cosine of the true anomaly (t) are computed
(5.2.36)
sin (t) =

1e
2
sin E(t)
1e cos E(t)
cos (t) =
cos E(t)e
1e cos E(t)
5.3. DISTURBED KEPLERIAN MOTION 33
Now the position of the satellite in the orbital system can be computed
(5.2.37)
_
_

_
_
= r(t)
_
_
cos (t)
sin (t)
0
_
_
r(t) =
p
1 +e cos (t)
The relation between the orbital system and CIS is given by
(5.2.38)
_
_
X
1
X
2
X
3
_
_
= R
3
()R
1
(i)R
3
()
_
_

_
_
5.3. Disturbed Keplerian Motion
The dierential equation for the undisturbed Keplerian motion X
0
(t) is homoge-
neous
(5.3.1)

X
0
+
GM
|X
0
|
3
X
0
= 0
The actual equation of motion is in contrast to (5.3.1)
(5.3.2)

X+
GM
|X|
3
X = F(X(t), t),
where F collects all disturbing forces acting on the satellite
(5.3.3) F = F
N
+F
M
+F
S
+F
T
+F
F
with
F
N
- nonspherical part of the Earths gravitational force
F
M
, F
S
- gravitational force of the Moon and Sun
F
N
- gravitational force of the tidal deformation of the Earth
F
F
- force of the atmospheric friction.
The right-hand-side of (5.3.2) is highly complex and prevents a solution in closed
form. For this reason an initial simplication is made. This simplication uses the
fact that the disturbing forces are much smaller than the Keplerian attraction force
|
GM
|X|
3
X| |F(X(t), t)|.
Therefore, the problem is only slightly changed by the approximation
F(X(t), t) F(X
0
(t), t) =: F
0
(t)
With this simplication the homogeneous problem (5.3.1) of the undisturbed Kep-
lerian motion changed into the corresponding inhomogeneous problem
(5.3.4)

X+
GM
|X|
3
X = F(t)
for the disturbed Keplerian problem. The solution of an inhomogeneous problem
with a known solution for the corresponding homogeneous equation can be con-
structed by the technique of the variation of the integration constants.
Before applying the variation of the constants method to (5.3.4), the essence of this
technique will be explained.
34 5. THEORY OF SATELLITE MOTION
5.3.1. Variation of the integration constants. The basic ideas of this
technique will be explained for the simple example of a scalar, linear dierential
equation of second order
(5.3.5) y

+py

+qy = f
The solution of (5.3.5) can be split o into the general solution y
h
of the correspond-
ing homogeneous problem and a special solution of the inhomogeneous problem y
i
(5.3.6) y = y
h
+y
i
The homogeneous equation
(5.3.7) y

+py

+qy = 0
has two linear independent fundamental solutions y
1
, y
2
and the general solution of
(5.3.7) is the linear span of these two fundamental solutions
y
h
= C
1
y
1
+C
2
y
2
For the unknown inhomogeneous solution y
i
, we use the ansatz
(5.3.8) y
i
= C
1
(t) y
1
(t) +C
2
(t) y
2
(t)
we try to vary the constants C
1
, C
2
of the homogeneous solution in such a way that
the inhomogeneous solution is generated. This ansatz motivates the name variation
of the constants for this technique.
The determination of the unknown functions is done by insertion of the ansatz
(5.3.8) into the inhomogeneous equation (5.3.5)
(5.3.9) (C
1
y
1
+C
2
y
2
)

+p (C
1
y
1
+C
2
y
2
)

+q (C
1
y
1
+C
2
y
2
) = f
Equation (5.3.9) forms one constraint for the two unknown functions C
1
, C
2
. In
order to get an unambiguous solution a second constraint
(5.3.10) C

1
y
1
+C

2
y
2
= 0
has to be added. Using (5.3.10), we obtain
(5.3.11)
y

i
= C

1
y
1
+C
1
y

1
+C

2
y
2
+C
2
y

2
= C
1
y

1
+C
2
y

2
y

i
= C

1
y

1
+C
1
y

1
+C

2
y

2
+C
2
y

2
The combination of (5.3.9) and (5.3.11) yields
(5.3.12) C
1
[y

1
+py

1
+qy
1
. .
0
] +C
2
[y

2
+py

2
+qy
2
. .
0
] +C

1
y

1
+C

2
y

2
= f
and we arrive at two linear equation for the derivatives C

1
, C

2
of the unknown
function C
1
, C
2
.
(5.3.13)
_
_
_
C

1
y
1
+C

2
y
2
= 0
C

1
y

1
+C

2
y

2
= f
The determinant of this equation

y
1
y
2
y

1
y

= y
1
y

2
y
2
y

1
5.4. VARIATION OF INTEGRATION CONSTANTS... 35
is identic to the Wronskian of y
1
, y
2
and therefore always dierent from zero. There-
fore (5.3.13) has the unique solution
(5.3.14)
C

1
=
fy
2
y
1
y

2
y
2
y

1
C

2
=
fy
1
y
1
y

2
y
2
y

1
Equation (5.3.14) is sometimes called disturbing equation since it describes the way
the inhomogeneity f disturbs the constants C
1
, C
2
of the homogeneous problem.
5.4. Variation of integration constants for disturbed orbit computation
The Keplerian orbit- the solution of the homogeneous problem-

X(t)
GM
|

X
0
(t)|
3

X
0
(t) = 0
has the general form
X
0
(t) = X
0
(t, u
1
, . . . , u
6
)
The parameters u
1
, . . . , u
6
are an arbitrary set of integration constants, e.g. the
Keplerian orbital elements. For the determination of the solution of the inhomoge-
neous equation

X+
GM
|X|
3
X = F(t)
the following ansatz is made
X(t) = X
0
(t, u
1
(t), . . . , u
6
(t)),
i.e. the disturbed orbit is represented as a sequence of Keplerian orbital ellipses,
which constantly change their shapes and their orientations in space.
These changing ellipses are usually called osculating orbital ellipses. The only prob-
lem which has to be solved is to determine the timely variation of the orbital con-
stants u
1
(t), . . . , u
6
(t).
In close analogy to (5.3.14), these variations are the solution of the so-called Gauss-
ian disturbing equations
(5.4.1)
da
dt
=
2
na
F,
X
0
M
)
de
dt
=
1e
2
na
2
e
F,
X
0
M
)

1e
2
na
2
e
F,
X
0

)
di
dt
=
1
na
2

1e
2
sin i
_
cos iF,
X
0

) F,
X
0

)
_
d
dt
=
cos i
na
2

1e
2
sin i
F,
X
0
i
) +

1e
2
na
2
e
F,
X
0
e
)
d
dt
=
1
na
2

1e
2
sin i
F,
X
0
i
)
dM
dt
= n
1e
2
na
2
e
F,
X
0
e
)
2
na
F,
X
0
a
)
The vector F is the total disturbing force acting on the satellite. It contains con-
servative forces- as the irregular gravity forces- as well as nonconservative forces as
the atmospheric friction.
The main components of the disturbing force F are of course the conservative
gravitational forces. Hence, there exist a potential V with
F = grad V
36 5. THEORY OF SATELLITE MOTION
This leads to a simplied form of the Gaussian disturbing equation (5.4.1), the
so-called Lagrangian disturbing equations:
(5.4.2)
da
dt
=
2
na
V
M
de
dt
=
1e
2
na
2
e
V
M

1e
2
na
2
e
V

di
dt
=
1
na
2

1e
2
sin i
_
cos i
V

_
d
dt
=
cos i
na
2

1e
2
sin i
V
i
+

1e
2
na
2
e
V
e
d
dt
=
1
na
2

1e
2
sin i
V
i
dM
dt
= n
1e
2
na
2
e
V
e

2
na
V
a
Equations (5.4.2) require the knowledge of the partial derivatives of the potential
with respect to the orbital elements a, e, i, , , M. Originally, the potential V is
a function of the position vector x in space and not of the orbital elements. Nev-
ertheless, these partial derivatives have not to be computed at arbitrary positions
in space but only along the reference-orbit X
0
(t) = X
0
(t, a, e, i, , , M). Inserting
this relation into the spherical harmonics expansion
V (r, , ) =
GM
R

l=2
l

m=0
_
R
r
_
l+1
P
lm
(sin ) [C
lm
cos m +S
lm
sin m]
yields a rather complicated series representation of V as a function of the orbital
elements.
(5.4.3) V =
GM
R

l=2
l

m=0
l

p=0

q=
_
R
a
_
l+1
F
lmp
(i) G
lpq
(e) S
lmpq
(, M, , )
Here the terms of the series are split o in
- the shape terms
_
R
r
_
l+1
F
lmp
(i)- the inclination function
G
lpq
(e)- the eccentricity function
- angular terms
S
lmpq
(, M, , )
Both the inclination function F
lmp
and the eccentricity function G
lpq
are series
expansion
G
lpq
(e) = X
l1,l2p
l2p+q
(e)
X
n,m
k
(e) =
1
(1 +
2
)
n+1

p=0

q=0
_
n +m+ 1
p
__
n m+ 1
q
_
()
p+q
J
km+pq
(ke)
with
=
e
1 +

1 e
2
5.4. VARIATION OF INTEGRATION CONSTANTS... 37
an J
j
being the Bessel-function of the rst kind. For the series expansion of the
inclination function we have
F
lmp
(i) = (1)
lm
2
(l +m)!
2
l
l!
(lp)
j
2

j=j
1
(1)
j
_
2p
j
__
2l 2p
l mj
_

_
cos
i
2
_
lm2p+2j
_
sin
i
2
_
lm2p2j
with
j
1
= max0, 2p l m , j
2
= maxl m, 2p
Fortunately, since the eccentricity e is very small for most geodetic satellites, the
series expansion can be truncated.
G
lp 0
= 1 +g
lp 0
e
2
2
+O(e
4
)
G
lp 1
= g
lp 1
e +O(e
3
)
G
lp 2
= g
lp 2
e
2
2
+O(e
4
)
with
g
lp 0
=
l+(4p3l)(l4p)
2
g
lp 1
=
3l4p+1
2
g
lp 1
=
4pl+1
2
g
lp 2
= (l p)(2l 3p +
5
2
) +
(l2p+2)
2
4
g
lp 2
= p(3p l +
5
2
) +
(l2p+2)
2
4
For indices [q[ > 2, the eccentric function can be neglected.
The remaining angular term S
lmpq
is dened by
S
lmpq
(, M, , ) = A
lm
cos
lmpq
+B
lm
sin
lmpq
,
with

lmpq
= (l 2p +q)M + (l 2p) +m( ),
= GAST
and the coecient A
lm
and B
lm
are related to the spherical harmonic coecients
C
lm
and S
lm
by
A
lm
=
_
C
lm
, l m even
S
lm
, l m odd
B
lm
=
_
S
lm
, l m even
C
lm
, l m odd.
Inserting Eq. (5.4.3) into the Lagrangian disturbing equations (5.4.2) yields
da
dt
=
2
na
GM
R

l=2
l

m=0
l

p=0

q=
_
R
a
_
l+1
F
lmp
(i) G
lpq
(e)
S
lmpq

lmpq
(l 2p +q)
38 5. THEORY OF SATELLITE MOTION
de
dt
=
GM
R

l=2
l

m=0
l

p=0

q=
_
R
a
_
l+1
F
lmp
(i) G
lpq
(e)
S
lmpq

lmpq

_
1 e
2
na
2
e
(l 2p +q)

1 e
2
na
2
e
(l 2p)
_
di
dt
=
GM
R

l=2
l

m=0
l

p=0

q=
_
R
a
_
l+1
F
lmp
(i) G
lpq
(e)
S
lmpq

lmpq

[cos i(l 2p) m]

1
na
2

1 e
2
sin i
d
dt
=
GM
R

l=2
l

m=0
l

p=0

q=
_
R
a
_
l+1
S
lmpq
_
cos i
na
2
(1 e
2
) sin i
F

lmp
(i)G
lpq
(e) +

1 e
2
na
2
e
F
lmp
(i)G

lpq
(e)
_
d
dt
=
GM
R

l=2
l

m=0
l

p=0

q=
_
R
a
_
l+1
S
lmpq
G
lmpq
(e)
F

lmp
(i)
na
2

1 e
2
sin i
dM
dt
= n
GM
R

l=2
l

m=0
l

p=0

q=
_
R
a
_
l+1
S
lmpq
F
lmp

_
1 e
2
na
2
e
G

lpq
(e) +
2(l + 1)
na
2
G
lpq
(e)
_
5.5. Integration of the disturbing equations
5.5.1. Orbit disturbances due to the Earths attening. The Earths
attening has the largest eect onto a satellite orbit. The attening is represented
by the term
V
2
=
GM
R
_
R
r
_
3
P
2
(sin ) C
20
in the spherical harmonics expansion of the Earths gravitational potential. It is
about one-thousand times larger than any of the other terms
V
2
> 1000
GM
R
_
R
r
_
l+1
P
lm
(sin )
_
C
lm
cos m
S
lm
sin m.
Therefore, it is useful to integrate as a rst step the Lagrangian disturbing equations
using the potential term V
2
only. Due to the dominance of V
2
, the solution will
represent the main part of the orbital disturbances.
First V
20
has to be represented as a function of the orbital elements.
V
2
=
GM
R
_
R
a
_
3 2

p=0
2

q=2
F
20p
G
2pq
S
20pq
As a next approximation we neglect all terms smaller than O(e). Since G
lpq
(e) =
O(e
|q|
) only the index q = 0 has to be considered. For reasons, which will be
5.5. INTEGRATION OF THE DISTURBING EQUATIONS 39
considered in a later section also the restriction to p = 1 is possible. The resulting
representation is
V
2
=
GM
R
_
R
a
_
3
F
201
G
210
[A
20
cos
2010
+B
20
sin
2010
]
=
GMR
2
a
3
F
201
G
210
C
20
=
GMR
2
a
3
_
3 sin
2 i
4

1
2
_ _
1 e
2
_

3
2
C
20
This results in
da
dt
=
2
na
V
2
M
= 0
de
dt
=
1 e
2
na
2
e
V
2
M

1 e
2
na
2
e
V
2

= 0
di
dt
=
1
na
2

1 e
2
sin i
_
cos i
V
2

V
2

_
= 0
d
dt
=
cos i
na
2

1 e
2
sin i
V
2
i

1 e
2
na
2
e
V
2
e
= n
_
R
a
_
2

3
2
cos i sin
i
4
cos
i
4
(1 e
2
)
2
sin i
C
20

3n
2
_
R
a
_
2
_
3 sin
2
i
4

1
2
_
2eC
20
(1 e
2
)
2
=
3
4
n
_
R
a
_
2
C
20
1 5 cos
2
i
(1 e
2
)
2
d
dt
= n
_
R
a
_
2
C
20
3
2
sin
i
4
cos
i
4
(1 e
2
)
2
sin i
=
3
2
nC
20
_
R
a
_
2
cos i
(1 e
2
)
2
dM
dt
= n
3
4
nC
20
_
R
a
_
2
3 cos
2
i 1
(1 e
2
)
3
2
The solution of these equations can immediately be given
a(t) = a
0
= const.
e(t) = e
0
= const.
i(t) = i
0
= const.
(t) =
0
+
3
4
n
0
C
20
_
R
a
_
2
1 5 cos i
0
(1 e
2
0
)
2
t
(t) =
0
+
3
2
n
0
C
20
_
R
a
_
2
cos i
0
(1 e
2
0
)
2
t
M(t) = n
0
t
3
4
n
0
C
20
_
R
a
_
2
3 cos
2
i
0
1
(1 e
2
0
)
3/2
t
40 5. THEORY OF SATELLITE MOTION
This means that the shape of the orbital ellipse and the inclination of the orbital
plane is not changed by the Earths attening. The attening causes only a pre-
cession of the orbital plane around the polar axis of the Earth and a rotation of
the orbital ellipse within the orbital plane. The rotation rates are dominantly in-
uenced by the ratio
R
a
, i.e. the closer the satellite is to the Earth the quicker the
precession of the orbital plane and the rotation of the apsidal lines.
The direction of precession and rotation depends on the orbital inclination. Since
C
20
< 0 the precession of the orbital plane is
retrograde for 0 i

2
prograde for

2
i
The line of apsides rotates
clockwise for i > i
0
clockwise for i < i
0
The critical inclination i
0
with stationary lines of apsides is given by
1 5 cos
2
i
0
= 0
leading to i
0
= 63

26

.
Figures (5.5.1) and (5.5.2) show the daily change of and as a function of the
inclination i for dierent orbital heights respectively.
Figure 5.5.1. nodal movement versus inclination
5.5. INTEGRATION OF THE DISTURBING EQUATIONS 41
Figure 5.5.2. apsidal line spine rate versus inclination
5.5.2. General solution of disturbing equations. The general form of the
Lagrangian disturbing equations is
da
dt
=
2
na
V
M
de
dt
=
1e
2
na
2
e
V
M

1e
2
na
2
e
V

di
dt
=
1
na
2

1e
2
sin i
_
cos i
V

_
d
dt
=
cos i
na
2

1e
2
sin i
V
i
+

1e
2
na
2
e
V
e
d
dt
=
1
na
2

1e
2
sin i
V
i
dM
dt
= n
1e
2
na
2
e
V
e

2
na
V
a
with
V =

l=2
l

m=0

l=0

q=
V
lmpq
and
V
lmpq
=
GM
R
_
R
a
_
l+1
F
lmp
(i)G
lmq
(e)S
lmpq
(, M, , )
The function S
lmpq
is dened by
S
lmpq
(, M, , ) = A
lm
cos
lmpq
+B
lm
sin
lmpq
,
and

lmpq
= (l 2p +q)M + (l 2p) +m( ).
42 5. THEORY OF SATELLITE MOTION
For the solution of the Lagrangian disturbing equations the superposition assump-
tion is made.
a(t) =

l=2
l

m=0

l=0

q=
a
lmpq
(t)
with a
lmpq
(t) solving
da
lmpq
dt
=
2
na
V
lmpq
M
The same assumption is made for the rest of the orbital elements. Hence, the
index-wise Lagrangian disturbing equations are
da
lmpq
dt
=
2
na
V
lmpq
M
=
2GM
Ra
_
R
a
_
l+1
F
lmp
G
lpq
S
lmpq

lmpq
(l 2p +q)
de
lmpq
dt
=
1 e
2
na
2
e
V
lmpq
M

1 e
2
na
2
e
V
lmpq

=
GM
na
2
e
_
R
a
_
l+1
F
lmp
G
lpq
S
lmpq

lmpq
_
(1 e
2
)(l 2p +q)
_
1 e
2
(l 2p)
_
di
lmpq
dt
=
1
na
2

1 e
2
sin i
_
cos i
V
lmpq

V
lmpq

_
=
GM
na
2

1 e
2
sin i
_
R
a
_
l+1
F
lmp
G
lpq
S
lmpq

lmpq
((l 2p) cos i m)
d
lmpq
dt
=
1
na
2

1 e
2
sin i
V
lmpq
i
=
GM
na
2

1 e
2
sin i
_
R
a
_
l+1
F

lmp
G
lpq
S
lmpq
d
lmpq
dt
=
cos i
na
2

1 e
2
sin i
V
lmpq
i
+

1 e
2
na
2
e
V
lmpq
e
=
GM
na
2
_
R
a
_
l+1
S
lmpq
_

1 e
2
e
F
lmp
G

lpq

cot i

1 e
2
F

lmp
G
lpq
_
dM
lmpq
dt
= n
1 e
2
na
2
e
V
lmpq
e

2
na
V
lmpq
a
= n
GM
na
2
_
R
a
_
l+1
F
lmp
S
lmpq
_
1 e
2
e
G

lpq
2a(l + 1)G
lpq
_
If additionally
,

M,

,

= const.
is assumed

lmpq
(t) =

lmpq
t

lmpq
= (l 2p) + (l 2p +q)

M +m(



)
5.5. INTEGRATION OF THE DISTURBING EQUATIONS 43
follows. Consequently, the time variable t can be uniquely expressed by the angular
quantity
lmpq
:
t =

lmpq

lmpq
Using the abbreviation

S
lmpq
=
_

0
S
lmpq
d
the solution of the index-wise Lagrangian disturbing equations can be given by
a
lmpq
=
_
t
0
da
lmpq
dt
dt =
_

0
da
lmpq
dt
1

lmpq
d
=
2GM
Ra
_
R
a
_
l+1
F
lmp
G
lpq
S
lmpq

lmpq
(l 2p +q)
e
lmpq
=
_
t
0
de
lmpq
dt
dt =
_

0
de
lmpq
dt
1

lmpq
d
=
GM
na
2
e
_
R
a
_
l+1
F
lmp
G
lpq
S
lmpq

lmpq
_
(1 e
2
)(l 2p +q)
_
1 e
2
(l 2p)
_
i
lmpq
=
_
t
0
di
lmpq
dt
dt =
_

0
di
lmpq
dt
1

lmpq
d
=
GM
na
2

1 e
2
sin i
_
R
a
_
l+1
F
lmp
G
lpq
S
lmpq

lmpq
((l 2p) cos i m)

lmpq
=
_
t
0
d
lmpq
dt
dt =
_

0
d
lmpq
dt
1

lmpq
d
=
GM
na
2

1 e
2
sin i
_
R
a
_
l+1
F

lmp
G
lpq

S
lmpq

lmpq

lmpq
=
_
t
0
d
lmpq
dt
dt =
_

0
d
lmpq
dt
1

lmpq
d
=
GM
na
2
_
R
a
_
l+1

S
lmpq

lmpq
_

1 e
2
e
F
lmp
G

lpq

cot i

1 e
2
F

lmp
G
lpq
_
M
lmpq
=
GM
na
2
_
R
a
_
l+1
F
lmp

S
lmpq

lmpq
_
1 e
2
e
G

lpq
2a(l + 1)G
lpq
_
These formulae can be used for a classication of disturbances.
44 5. THEORY OF SATELLITE MOTION
5.5.3. Classication of orbital disturbances. Orbital disturbances are clas-
sied in four classes:
secular disturbances, i.e. non-periodic orbital elements changes,
long-periodic disturbances, i.e. changes of the orbital elements with a
period longer than the orbital period T of the satellite,
short-periodic disturbances, i.e. changes of the orbital elements with a
period shorter than the orbital period T of the satellite,
resonant disturbances.
Since the time variability of the orbital disturbances a
lmpq
, e
lmpq
, . . . are solemnly
generated by the term S
lmpq
, the behavior of S
lmpq
decides the periodicity of the
orbital disturbances.
5.5.3.1. Secular disturbances. A non-periodic disturbance can only be gener-
ated if S
lmpq
= const. Therefore

lmpq
= 0 has to be valid. This leads to the
condition
l 2p = l 2p +q = m = 0
the condition m = 0 leads to the conclusion that secular disturbances can only be
generated by zonal harmonic coecients C
l0
.
In the investigations of disturbances, caused by C
20
, the largest disturbances were
in the focus. Secular disturbances are larger than any other disturbances if only
enough time has elapsed. The condition for secular disturbances is
0 = l 2p = 2 2p = 0 = p = 1
This was the reason for setting p = 1 in Section. (5.5.1).
5.5.3.2. Long-periodic disturbances. In order to generate long-periodic distur-
bances the angular velocity

lmpq
has to be small. Recall that

lmpq
= (l 2p) + (l 2p +q)

M +m(



)
holds. Each term in

lmpq
produces a dierent period
- period of several days,

M - orbital period T of the satellite,

,

- about daily period.
In order to be long-periodic, the

M and (



) have to vanish in

lmpq
. This
leads to the conditions
(l 2p +q) = m = 0
l 2p ,= 0
As in the case of the secular disturbances, the long-periodic disturbances are gener-
ated by zonal harmonics. Because of l 2p ,= 0 only odd zonal harmonics C
2k+1,0
are responsible for long-periodic disturbances.
5.5.3.3. Short-periodic disturbances. Completely analog to the long-periodic
case, one can reasons that short-periodic eects are generated, if

lmpq
large
hold. This is the case if
(l 2p +q) ,= 0 and/or m ,= 0
holds. This means short-periodic disturbance are produced by tesseral harmonics.
5.5. INTEGRATION OF THE DISTURBING EQUATIONS 45
5.5.3.4. Resonant disturbances. The solution of the index-wise Lagrangian equa-
tions always have the term

lmpq
in the denominator. Therefore, the disturbances
get particularly large if

lmpq
is close to zero. Since in general and

are very
small this is the case if
(l 2p +q)

M = m

holds. This is the case if

M
=
(l 2p +q)
m
holds, i.e. ratio between the length of a day and the orbital period T of the satellite
is a rational number.
Dierent kinds of orbital disturbances act dierently on the orbital elements. Table
(1) shows which elements are secular, long-periodic or short-periodic disturbances.
Table 1. Characteristics of disturbances in the Keplerian elements
element secular long-periodic short-periodic
a - -
e -
i -


M
CHAPTER 6
Numerical Orbit Integration
Analytical methods of orbit integration provide a good insight in the nature of
the satellite motion but they are of limited accuracy. For high-accuracy application
the equation of motion
(6.0.1) x = F(t, x, x) , x(0) = x
0
, x(0) = x
0
Here numerical integrators for initial value problem of the rst order
(6.0.2) y

(x) = f(x, y(x)), y(x

0
) = y
0
will be discussed. Before, an application of the numerical integrators to the equa-
tion of motion (Eq. 6.0.1) becomes possible, the second order problem has to be
transformed into an equivalent problem of the rst order. This can be accomplished
by the substitution
(6.0.3) y

= (y

1
, y

2
) = (x

, x

)
yielding
(6.0.4) y =
_
y
2
F(t, y
1
, y
2
)
_
, y

(0) = (x

0
, x

0
)
The basic idea of numerical integration is not to look for a solution of Eq. (6.0.4)
for all t > 0 but only for discrete values t
k
= k h.
Two dierent types of numerical integrators can be distinguished: Integrators
which uses only the known solution y(t
k
) for the computation of the solution
at the next point y(t
k+1
) and those which also use the solution at earlier points
y(t
k
), y(t
k1
), y(t
k2
), . . . . The rst class of integrators is called single-step meth-
ods and the latter multi-step methods.
6.1. Single-step methods
The basic idea of a single-step methods is to replace the dierential equation
(6.1.1) y

(x) = f(x, y(x)), y(x

0
) = y
0
by an equivalent integral equation
(6.1.2) y(x
k+1
) = y(x
k
) +
_
x
k+1
x
k
y

(x)dx = y(x
k
) +
_
x
k+1
x
k
f(x, y(x)) dx
The integral in the new equation can be evaluated using a quadrature formula.
Depending on the kind of quadrature formula which is applied, dierent integrators
are generated. The simplest quadrature formula is the rectangular rule:
(6.1.3)
_
b
a
f(x)dx = (b a)f(a)
If the rectangular rule is applied to the integral equation (6.1.2) the following inte-
gration formula is obtained:
(6.1.4) y(x
k+1
) = y(x
k
) +hf(x
k
, y(x
k
)), k = 0, 1, . . .
47
48 6. NUMERICAL ORBIT INTEGRATION
This integration method is usually called Euler - Cauchy method or sometimes also
polygon method .
Euler-Cauchys method is an explicit single step method, meaning that the new
approximation y(x
k+1
) can be computed directly from the known approximation
y(x
k
) , without solving an equation. The most striking disadvantage of this method
is the increase of the integration error with increasing step number k. The reason for
this undesired eect is the exceedingly simple structure of the underlying quadrature
formula.
Using a better quadrature formula, e.g. the trapezoid rule
(6.1.5)
_
b
a
f(x)dx =
b a
2
(f(a) +f(b))
generates the so-called Heun formula:
(6.1.6) y(x
k+1
) = y(x
k
) +
h
2
(f(x
k
, y(x
k
)) +f(x
k+1
, y(x
k+1
)), k = 0, 1, . . .
The error behavior of Heuns formula is noticeably better than that of Euler-Cauchy
method. On the other hand, heunss formula is an implicit method, since the
unknown new approximation y(x
k+1
) occurs both on the left- and right-hand-side of
equation (6.1.6). Therefore, it cannot be computed directly but via an intermediate
equation solution. The following form is obtained by implementing this equation
solution into Heuns formula.
(6.1.7)
0
= y(x
k
) +hf(x
k
, y(x
k
)) Euler Cauchy approximate solution
(6.1.8)
j+1
= y(x
k
) +
h
2
(f(x
k
, y(x
k
)) +f(x
k+1
,
j
)), j = 0, . . . , N
(6.1.9) y(x
k+1
) :=
N+1
k = 0, 1, . . .
After computing an initial approximation
0
for instance by one step of Euler-
Cauchys method, the solution of (6.1.6) is computed by N steps of the simple
xed-point iteration. In general the number of steps N is between 2 and 6. After
terminating the iteration, the last iterate,
N+1
, is chosen as the new approximation
of y(x
k+1
).
The interposed iteration in implicit integration methods is very in commodiously.
Therefore, other methods to overcome this obstacle are welcome. Stopping the
iteration after one single step is the simplest possibility.
:= y(x
k
) +hf(x
k
, y(x
k
)), (6.1.10)
y(x
k+1
) := y(x
k
) +
h
2
(f(x
k
, y(x
k
)) +f(x
k+1
, )) (6.1.11)
If this relation is rewritten in the following form
K
1
:= f(x
k
, y(x
k
)) (6.1.12)
K
2
:= f(x
k+1
, y(x
k
) +hK
1
) (6.1.13)
y(x
k+1
) = y(x
k
) +
h
2
(K
1
+K
2
) (6.1.14)
the simplest version of the so-called Runge-Kutta method is obtained. The classical
Runge-Kuttta method is generated if , instead of the trapezoid rule, Simpsons rule
is applied:
(6.1.15)
_
b
a
f(x)dx =
h
6
(f(a) + 4f(
a +b
2
) +f(b)), h = b a
6.2. LINEAR MULTI-STEP PROCEDURE 49
Consequently equation(6.1.2) can be recast into
(6.1.16)
y(x
k+1
) = y(x
k
)+
h
6
_
f(x
k
, y(x
k
)) + 2f(x
k+
1
2
, y(x
k+
1
2
)) + 2f(x
k+
1
2
, y(x
k+
1
2
)) +f(x
k+1
, y(x
k+1
)
_
.
The unknown values y(x
k+
1
2
), y(x
k+1
) are computed step-by-step
K
1
:= f(x
k
, y(x
k
)) (6.1.17)
K
2
:= f(x
k
+
h
2
, y(x
k
) +
h
2
K
1
) (6.1.18)
K
3
:= f(x
k
+
h
2
, y(x
k
) +
h
2
K
2
) (6.1.19)
K
4
:= f(x
k+1
, y(x
k
) +hK
3
) (6.1.20)
and yield the classical notation of Runge - Kutta - method:
(6.1.21) y(x
k+1
) = y(x
k
) +
h
6
(K
1
+ 2K
2
+ 2K
3
+K
4
)
All integration methods discussed so far, have the structure
(6.1.22) y(x
k+1
) = y(x
k
) +h(y(x
k
))
i.e. the increment function depends only on the actual approximation y(x
k
)
and not on earlier approximations y(x
k1
), y(x
k2
), . . .. This structure gave the
previously discussed group of integration methods the name single-step methods.
Single-step methods have the advantage, that a change of the integration step-size
h
k
= x
k+1
x
k
is possible in each integration step. They have the disadvantage
that in every step the function f has to be evaluated several times.
These properties are complementary to those of the so-called multi-step methods,
which will be discussed in the following section.
6.2. Linear Multi-step procedure
For the construction of multi-step methods the knowledge of previous values y(x
k
),
y(x
k1
), . . . , y(x
kq
) is supposed. These values can be computed by an initial
computation by q-steps of a single-step method. Based on these known values
y(x
k
), y(x
k1
), . . . , y(x
kq
) the function f(x, y(x)) is extrapolated onto the interval
[x
k
, x
k+1
]. This extrapolation is inserted into the integral equation (6.1.2) and the
integral is evaluated by a quadrature formula. If the extrapolation is carried out
by a polynomial
(6.2.1) f(x, y(x)) =
q

n=0
a
n
x
n
,
the integration even can be carried out in a closed form and the multi-step method
is generated.
The simplest formula, which can be generated in this way is a-two-step formula.
The extrapolation polynomial is the simple lagrangian interpolation polynomial
with two nodes. Supposed the values y(x
k
), y(x
k1
) are known the Lagrangian
interpolation polynomial is
(6.2.2) f(x, y(x))
x x
k1
x
k
x
k1
f(x
k
, y(x
k
)) +
x x
k
x
k1
x
k
f(x
k1
, y(x
k1
))
50 6. NUMERICAL ORBIT INTEGRATION
Table 1. Multi-step integration coecients

n
q
0
1
2
3
0 1 2 3
1
3/2 -1/2
23/12 -16/12 5/12
55/24 -59/24 37/24 -9/24
Inserting this polynomial into integral equation (6.1.2) yields
y(x
k+1
) = y(x
k
) +
1
h
_
x
k+1
x
k
_
(x x
k1
)f(x
k
, y(x
k
)) (x x
k
)f(x
k1
, y(x
k1
))
_
dx
= y(x
k
) +
1
h
_
(x x
k1
)
2
2
f(x
k
, y(x
k
))
(x x
k
)
2
2
f(x
k1
, y(x
k1
))
_
x
k+1
x
k
= y(x
k
) +
1
h
_
3h
2
2
f(x
k
, y(x
k
))
h
2
2
f(x
k1
, y(x
k1
))
_
= y(x
k
) +h(
3
2
f(x
k
, y(x
k
))
1
2
f(x
k1
, y(x
k1
)))
This procedure can be repeated also for dierent step-numbers q leading to multi-
step methods of the structure
(6.2.3) y(x
k+1
) = y(x
k
) +h
q

n=0

n
f(x
kq
, y(x
kq
)
Of course, the coecients
n
depend on the step-number q. The relation between
q and the corresponding coecients
i
is compiled in Table (1).
The advantage of the multi-step methods is that per step only one evaluation of
the function f is necessary. The disadvantages are
a start computation with a single-step method is necessary
it is dicult to change the step size h.
Bibliography
[1] Egorov Y. V. ; Shubin M. A. :Partial Dierential Equations I, Springer-Verlag Berlin , 1991.
[2] Hartmann G. K. ; Leitinger R. :Range errors due to ionospheric and tropospheric eects for
signal frequencies above 100 MHz, Bull. geod. 58, 109-136, 1984.
[3] Hopeld, H. :Two-quartic tropospheric refractivity prole for correcting satellite data, JGR
74, No. 18, 1969.
[4] McCarthy, D. D. (ed.) :IERS Conventions, IERS Technical Note 21, Observatoire de Paris,
Paris , 1996.
51