Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

Contents
1 Introduction

2 Definitions
2.1 The Potential and the Geoid . . . . . . . . . . . .
2.2 The Height Anomaly . . . . . . . . . . . . . . . . .
2.3 The Gravity Disturbance . . . . . . . . . . . . . .
2.4 The Gravity Anomaly . . . . . . . . . . . . . . . .
2.4.1 The Classical Definition . . . . . . . . . . .
2.4.2 The Modern Definition . . . . . . . . . . . .
2.4.3 The Topography-Reduced Gravity Anomaly

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

2
2
5
6
7
7
7
7

3 Approximation and Calculation

3.1 The Geoid . . . . . . . . . . . . . . . . . . . . . . .
3.2 The Height Anomaly . . . . . . . . . . . . . . . . .
3.3 The Difference: Geoid - Height Anomaly . . . . . .
3.4 The Gravity Disturbance . . . . . . . . . . . . . .
3.5 The Gravity Anomaly . . . . . . . . . . . . . . . .
3.5.1 The Classical Gravity Anomaly . . . . . . .
3.5.2 The Modern Gravity Anomaly . . . . . . .
3.5.3 The Topography-Reduced Gravity Anomaly

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

8
8
13
14
14
15
15
16
17

4 Calculation from Spherical Harmonics

4.1 Spherical Harmonics and the Gravity Field
4.2 The Geoid . . . . . . . . . . . . . . . . . . .
4.3 The Height Anomaly . . . . . . . . . . . . .
4.4 The Gravity Disturbance . . . . . . . . . .
4.5 The Gravity Anomaly . . . . . . . . . . . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

17
17
21
22
22
23

5 Practical calculations using the model EIGEN-GL04C

5.1 Geoid and Height Anomaly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Gravity Disturbance and Gravity Anomaly . . . . . . . . . . . . . . . . . . . . . . . . . .

24
24
24

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

Introduction

The intention of this article is to present the definitions of different functionals of the Earths gravity
field and possibilities for their approximative calculation from a mathematical representation of the outer
potential. In history this topic has usually been treated in connection with the boundary value problems
of geodesy, i.e. starting from measurements at the Earths surface and their use to derive a mathematical
representation of the geopotential.
Nowadays global gravity field models, mainly derived from satellite measurements, become more and
more detailed and accurate and, additionally, the global topography can be determined by modern satellite
methods independently from the gravity field. On the one hand the accuracy of these gravity field
models has to be evaluated and on the other hand they should be combined with classical (e.g. gravity
anomalies) or recent (e.g. GPS-levelling-derived or altimetry-derived geoid heights) data. Furthermore,
an important task of geodesy is to make the gravity field functionals available to other geosciences. For
all these purposes it is necessary to calculate the corresponding functionals as accurately as possible
or, at least, with a well-defined accuracy from a given global gravity field model and, if required, with
simultaneous consideration of the topography model.
We will start from the potential, formulate the definition of some functionals and derive the formulas
for the calculation. In doing so we assume that the Earths gravity potential is known outside the masses,
the normal potential outside the ellipsoid and that mathematical representations are available for both.
Here we neglect time variations and deal with the stationary part of the potential only.
Approximate calculation formulas with different accuracies are formulated and specified for the case
that the mathematical representation of the potential is in terms of spherical harmonics. The accuracies
of the formulas are demonstrated by practical calculations using the gravity field model EIGEN-GL04C
(Forste et al., 2006).
More or less, what is compiled here is well-known in physical geodesy but distributed over a lot of
articles and books which are not cited here. In the first instance this text is targeted at non-geodesists
and it should be stand-alone readable.
Textbooks for further study of physical geodesy are (Heiskanen & Moritz, 1967; Pick et al., 1973;
Vancek & Krakiwsky, 1982; Torge, 1991; Moritz, 1989; Hofmann-Wellenhof & Moritz, 2005).

2
2.1

Definitions
The Potential and the Geoid

As it is well-known, according to Newtons law of gravitation, the potential Wa of an attractive body

with mass density is the integral (written in cartesian coordinates x, y, z)
ZZZ

(x0 , y 0 , z 0 )

Wa (x, y, z) = G
v

(x x0 )2 + (y y 0 )2 + (z z 0 )2

dx0 dy 0 dz 0

(1)

0
0
0
over the volume v of the body,
p where G is the Newtonian gravitational constant, and dv = dx dy dz is
0
2
0
2
0
2
the element of volume. For (x x ) + (y y ) + (z z ) the potential Wa behaves like the
potential of a point mass located at the bodies centre of mass with the total mass of the body. It can be
shown that Wa satisfies Poissons equation

2 Wa = 4G

(2)

where is the Nabla operator and 2 is called the Laplace operator (e.g. Bronshtein et al., 2004).
Outside the masses the density is zero and Wa satisfies Laplaces equation
2 W a = 0
2

(3)

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

thus Wa is a harmonic function in empty space (e.g. Blakely, 1995).

On the rotating Earth, additionally to the attracting force, also the centrifugal force is acting which
can be described by its (non-harmonic) centrifugal potential
1 2 2
dz
(4)
2
p
where is the angular velocity of the Earth and dz = x2 + y 2 is the distance to the rotational (z-) axis.
Hence, the potential W associated with the rotating Earth (e.g. in an Earth-fixed rotating coordinate
system) is the sum of the attraction potential Wa and the centrifugal potential
(x, y, z) =

W = Wa +

(5)

The associated force vector ~g acting on a unit mass, the gravity vector, is the gradient of the potential
~g = W

(6)

g = |W |

(7)

and the magnitude

is called gravity. Potentials can be described (and intuitively visualised) by its equipotential surfaces.
From the theory of harmonic functions it is known, that the knowledge of one equipotential surface is
sufficient to define the whole harmonic function outside this surface.
For the Earth one equipotential surface is of particular importance: the geoid. Among all equipotential
surfaces, the geoid is the one which coincides with the undisturbed sea surface (i.e. sea in static equilibrium) and its fictitious continuation below the continents as sketched in Fig. 1 (e.g. Vancek & Christou,
1994, Vancek & Krakiwsky, 1982 or Hofmann-Wellenhof & Moritz, 2005). Being an equipotential sur-

topography
tor

y vec
gravit

H
ht

geoid

W = Uo

ellipsoid
U = Uo

Figure 1: The ellipsoid, the geoid and the topography

face, the geoid is a surface to which the force of gravity is everywhere perpendicular (but not equal in
magnitude!). To define the geoid surface in space, simply the correct value W0 of the potential has to be
chosen:
W (x, y, z) = W0 = constant
(8)
As usual we split the potential W into the normal potential U and the disturbing potential T
W (x, y, z) = U (x, y, z) + T (x, y, z)
3

(9)

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

and define shape and strengths of the normal potential as follows: (a) The equipotential surfaces
(U (x, y, z) = constant) of the normal potential should have the shapes of ellipsoids of revolution and (b)
the equipotential surface for which holds U (x, y, z) = W0 (see eq. 8) should approximate the geoid, i.e.
the undisturbed sea surface, as good as possible (i.e. in a least squares fit sense). It is advantageous to
define ellipsoidal coordinates (h, , ) with respect to this level ellipsoid U (h = 0) = U0 = W0 , where h
is the height above ellipsoid (measured along the ellipsoidal normal), is the ellipsoidal longitude and
the ellipsoidal latitude. Thus eq. (9) writes (note that the normal potential U does not depend on ):
W (h, , ) = U (h, ) + T (h, , )

(10)

and the geoid, in ellipsoidal coordinates, is the equipotential surface for which holds

W h = N (, ), , = U (h = 0), = U0

(11)

where N (, ) is the usual representation of the geoid as heights N with respect to the ellipsoid (U = U0 )
as a function of the coordinates and . Thus N are the undulations of the geoidal surface with respect
to the ellipsoid. This geometrical ellipsoid together with the normal ellipsoidal potential is called Geodetic
Reference System (e.g. NIMA, 2000 or Moritz, 1980). Now, with the ellipsoid and the geoid, we have two
reference surfaces with respect to which the height of a point can be given. We will denote the height
of the Earths surface, i.e. the height of the topography, with respect to the ellipsoid by ht , and with
respect to the geoid by H, hence it is (see fig. 1):
ht (, ) = N (, ) + H(, )

(12)

Here H is assumed to be measured along the ellipsoidal normal and not along the real plumb line, hence
it is not exactly the orthometric height. A discussion of this problem can be found in (Jekeli, 2000).
Like the potential W (eq. 5) the normal potential also consists of an attractive part Ua and the
centrifugal potential
U = Ua +
(13)
and obviously, the disturbing potential
T (h, , ) = Wa (h, , ) Ua (h, )

(14)

does not contain the centrifugal potential and is harmonic outside the masses. The gradient of the normal
potential
~ = U
(15)
is called normal gravity vector and the magnitude
= |U |

(16)

is the normal gravity.

For functions generated by mass distributions like Wa from eq. (1), which are harmonic outside the
masses, there are harmonic or analytic continuations Wac which are equal to Wa outside the masses and
are (unlike Wa ) also harmonic inside the generating masses. But the domain where Wa is harmonic, i.e.
satisfies Laplaces equation (eq. 3), can not be extended completely into its generating masses because
there must be singularities somewhere to generate the potential, at least, as is well known, at one point
at the centre if the mass distribution is spherically symmetric. How these singularities look like (point-,
line-, or surface-singularities) and how they are distributed depends on the structure of the function Wa
outside the masses, i.e. (due to eq. 1) on the density distribution of the masses. Generally one can say
that these singularities are deeper, e.g. closer to the centre of mass, if the potential Wa is smoother.
For further study of this topic see (Zidarov, 1990) or (Moritz, 1989).
4

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

Due to the fact that the height H = ht N of the topography with respect to the geoid is small
compared to the mean radius of the Earth and that in practise the spatial resolution (i.e. the roughness)
of the approximative model for the potential Wa will be limited (e.g. finite number of coefficients or
finite number of sampling points), we expect that the singularities of the downward continuation of Wa
lie deeper than the geoid and assume that Wac exists without singularities down to the geoid so that we
can define (ht is the ellipsoidal height of the Earths surface, see eq. 12):
Wac (h, , ) =
2 Wac =
c
W (h, , ) =

Wa (h, , )
for h ht
0
for h min(N, ht )
Wac (h, , ) + (h, )

(17)

However, this can not be guaranteed and has to be verified, at least numerically, in practical applications.
From its definition the normal potential Ua is harmonic outside the normal ellipsoid and it is known
that a harmonic downward continuation Uac exists down to a singular disk in the centre of the flattened
rotational ellipsoid (e.g. Zidarov, 1990). Thus, downward continuation of the normal potential is no
problem and we can define
Uac (h, ) = Ua (h, )
for
2 Uac = 0
for
U c (h, ) = Uac (h, ) + (h, )

h0
h min(N, 0, ht )

(18)

and hence
T c (h, , ) = Wac (h, , ) Uac (h, )
2 T c = 0
for

2.2

h min(N, ht )

(19)

The Height Anomaly

The height anomaly (, ), the well known approximation of the geoid undulation according to Molodenskys theory, can be defined by the distance from the Earths surface to the point where the normal
potential U has the same value as the geopotential W at the Earths surface (Molodensky et al., 1962;
Hofmann-Wellenhof & Moritz, 2005; Moritz, 1989):
W (ht , , ) = U (ht , , )

(20)

where ht is the ellipsoidal height of the Earths surface (eq. 12). An illustration of the geometrical
situation is given in fig. (2). The surface with the height = (, ) with respect to the ellipsoid is often
called quasigeoid (not shown in fig. 2) and the surface ht is called telluroid. It should be emphasised,
that the quasigeoid has no physical meaning but is an approximation of the geoid as we will see. In areas
where ht = N (or H = 0) i.e. over sea, the quasigeoid coincides with the geoid as can be seen easily from
the definition in eq. (20) if we use eq. (12):
W (N + H, , ) = U (N + H , , )

(21)

W (N, , ) = U (N , )

(22)

set H = 0 and get

and use eq. (11), the definition of the geoid to write
U (0, ) = U (N , )

(23)

from which follows

N =

for
5

H=0

(24)

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

topography

tial
oten ces
p
i
u
eq surfa

W(ht ) =
U(ht )

ht

N =

ht

telluroid

if

ht = N

equipotential surfaces

W = Uo geoid
U = Uo ellipsoid

Figure 2: The ellipsoid, the geoid, and the height anomaly

In the history of geodesy the great importance of the height anomaly was that it can be calculated
from gravity measurements carried out at the Earths surface without knowledge of the potential inside
the masses, i.e. without any hypothesis about the mass densities.
The definition of eq. (20) is not restricted to heights h = ht on the Earths surface, thus a generalised
height anomaly g = g (h, , ) for arbitrary heights h can be defined by:
W (h, , ) = U (h g , )

2.3

(25)

The Gravity Disturbance

The gradient of the disturbing potential T is called the gravity disturbance vector and is usually denoted
~
by g:
~
g(h,
, ) = T (h, , ) = W (h, , ) U (h, )
(26)
The gravity disturbance g is not the magnitude of the gravity disturbance vector (as one could guess)
but defined as the difference of the magnitudes (Hofmann-Wellenhof & Moritz, 2005):

g(h, , ) = W (h, , ) U (h, )

(27)
In principle, herewith g is defined for any height h if the potentials W and U are defined there. Additionally, with the downward continuations Wac and Uac (eqs. 17 and 18), we can define a harmonic
downward continued gravity disturbance

(28)
g c (h, , ) = W c (h, , ) U c (h, )
With the notations from eqs. (7) and (16) we can write the gravity disturbance in its common form:
g(h, , ) = g(h, , ) (h, )

(29)

The reason for this definition is the practical measurement process, where the gravimeter measures only
|W |, the magnitude of the gravity, and not the direction of the plumb line.
6

Scientific Technical Report 09/02

2.4

Deutsches GeoForschungsZentrum GFZ

The Gravity Anomaly

The term gravity anomaly is used with numerous different meanings in geodesy and geophysics and,
moreover, there are different practical realisations (cf. Hackney & Featherstone, 2003). Here we will
confine ourselves to the classical free air gravity anomaly, to the gravity anomaly according to Molodenskys
theory and to the topography-reduced gravity anomaly.
2.4.1

The Classical Definition

The classical (historical) definition in geodesy is the following (cf. Hofmann-Wellenhof & Moritz, 2005):
The gravity anomaly gcl (subscript cl stands for classical) is the magnitude of the downward
continued gravity |W c | (eq. 17) onto the geoid minus the normal gravity |U | on the ellipsoid at the
same ellipsoidal longitude and latitude :

gcl (, ) = W c (N, , ) U (0, )

(30)
The origin of this definition is the (historical) geodetic practise where the altitude of the gravity measurement was known only with respect to the geoid from levelling but not with respect to the ellipsoid.
The geoid height N was unknown and should be determined just by these measurements. The classical
formulation of this problem is the Stokes integral (e.g. Hofmann-Wellenhof & Moritz, 2005; Martinec,
1998). For this purpose the measured gravity |W (ht , , )| has to be reduced somehow down onto the
geoid and the exact way to do so is the harmonic downward continuation of the attraction potential Wa
(eq. 17). This is the reason for the definition of the classical gravity anomaly in eq. (30). In practise the
so-called free air reduction has been or is used to get |W c (N, , )| approximately. Thus the classical
gravity anomaly depends on longitude and latitude only and is not a function in space.
2.4.2

The Modern Definition

The generalised gravity anomaly g according to Molodenskys theory (Molodensky et al., 1962;
Hofmann-Wellenhof & Moritz, 2005; Moritz, 1989) is the magnitude of the gravity at a given point
(h, , ) minus the normal gravity at the same ellipsoidal longitude and latitude but at the ellipsoidal
height h g , where g is the generalised height anomaly from definition (25):

g(h, , ) = W (h, , ) U (h g , ),

for

h ht

(31)

or in its common form:

g(h, , ) = g(h, , ) (h g , )

(32)

Here the height h is assumed on or outside the Earths surface, i.e. h ht , hence with this definition
the gravity anomaly is a function in the space outside the masses. The advantage of this definition is
that the measured gravity |W | at the Earths surface can be used without downward continuation or
any reduction. If geodesists nowadays speak about gravity anomalies, they usually have in mind this
definition with h = ht , i.e. on the Earths surface.
2.4.3

The Topography-Reduced Gravity Anomaly

For many purposes a functional of the gravitational potential is needed which is the difference between the
real gravity and the gravity of the reference potential and which, additionally, does not contain the effect
of the topographical masses above the geoid. The well-known Bouguer anomaly or Refined Bouguer
anomaly (e.g. Hofmann-Wellenhof & Moritz, 2005) are commonly used in this connection. However,
they are defined by reduction formulas and not as functionals of the potential. The problems arising
when using the concepts of the Bouguer plate or the Bouguer shell are discussed in (Vancek et al., 2001)
and (Vancek et al., 2004).
7

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

Thus, let us define the gravity potential of the topography Vt , i.e. the potential induced by all masses
lying above the geoid. Analogously to eq. (27), we can now define a gravity disturbance gtr which does
not contain the gravity effect of the topography:

gtr (h, , ) = W (h, , ) Vt (h, , ) U (h, )

(33)

and, analogously to eq. (31), a topography-reduced gravity anomaly gtr :

gtr (h, , ) = W (h, , ) Vt (h, , ) U (h , )

(34)

where, consequently, W Vt is the gravity potential of the Earth without the masses above the geoid.
The difficulty here is, that the potential Vt (or its functionals) cannot be measured directly but can
only be calculated approximately by using a digital terrain model of the whole Earth and, moreover, a
hypothesis about the density distribution of the masses.
Approximate realisations of such anomalies are mainly used in geophysics and geology because they
show the effects of different rock densities of the subsurface. If geophysicists or geologists speak about
gravity anomalies they usually have in mind this type of anomalies.

3
3.1

Approximation and Calculation

The Geoid

As one can see from the definition in eq. (8) or eq. (11), the calculation of the geoid is the (iterative)
search of all points in space which have the same gravity potential W = W0 = U0 . Let us assume that the
geopotential W (h, , ) is known also inside the masses and Ni (, ) is a known approximative value for
the exact geoid height N (, ) (e.g. as result of the i-th step of an iterative procedure). Here we should
have in mind that the representation N of the geoidal surface is with respect to the normal ellipsoid
which already is a good approximation of the geoid in the sense that the biggest deviations of the geoid
from the ellipsoid with respect to its semi-major axis is in the order of 105 .
The difference W (N ) W (Ni ) for the coordinates and is (approximately):

W
W (N ) W (Ni ) (N Ni )
(35)
h h=Ni
The ellipsoidal elevation h is taken along the ellipsoidal normal which is given by the negative direction
U
W
of the gradient of the normal potential |U
| . Thus the partial derivative h can be represented by the
normal component of the gradient W , i.e. by the projection onto the normal plumb line direction

U
W
=

W
(36)
h
|U |
or, because the directions of W and U nearly coincide, by:

W
W

W
= W

h
|W |

(37)

where h~a | ~bi denotes the scalar product of the vectors ~a and ~b and, if ~a has the unit length as in eqs. (36)
and (37), the projection of ~b onto the direction of ~a. By replacing W (N ) by U0 according to eq. (11) and
with the notation g = |W | (eq. 7) we can write
U0 W (Ni ) g(Ni ) (N Ni )
8

(38)

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

for eq. (35) and thus the geoid height N can (approximately) be calculated by
N Ni +

1
W (Ni ) U0
g(Ni )

(39)

and the reasons for instead of = are the linearisation in eq. (35) and the approximation in eq. (37).
That means, if the gravity potential W is known also inside the topographic masses, eq. (39) can be used
to calculate the geoid iteratively with arbitrary accuracy for each point (, ):

1
W (Ni , , ) U0
g(Ni , , )

Ni+1 (, ) = Ni (, ) +

(40)

provided that we have an appropriate starting value for the iteration and the iteration converges. Replacing the gravity g(Ni ) (eq. 7) by the normal gravity (0) (eq. 16) will not change the behaviour of
this iteration, because each step will be scaled only by a factor of (1 g/), which is in the order of 104
or smaller. So we write:
Ni+1 (, ) = Ni (, ) +

1
W (Ni , , ) U0
(0, )

(41)

With i = 0 and N0 = 0 in eq. (41) we get:

N1 (, ) =

1
W (0, , ) U0
(0, )

(42)

and with W (0) = U0 + T (0) (eq. 10) we finally have

N1 (, ) =

T (0, , )
(0, )

(43)

as a first approximate value for N (, ) which is the well-known Bruns formula (e.g. Hofmann-Wellenhof
& Moritz, 2005).
To get an estimation of the difference N2 N1 from eq. (41) we write:

1
W (N1 , , ) U0
(0, )

(44)

1
U (N1 , , ) + T (N1 , , ) U0
(0, )

(45)

N2 (, ) N1 (, ) =
and replace again W by U + T (eq. 10) and get:
N2 (, ) N1 (, ) =
With the linearisation

U
U (N1 ) U (0) + N1
h h=0

and the notation (cf. eq. 16)

= U =

we get
N2 (, ) N1 (, )

U
U
U =

|U |
h

1
(0, )N1 + T (N1 , , )
(0, )

(46)

(47)

(48)

Replacing N1 on the right hand side by Bruns formula (eq. 43) we get
N2 (, ) N1 (, )

T (N1 , , ) T (0, , )
(0, )
9

(49)

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

for the difference, and for N2 :

N2 (, )

T (N1 , , )
(0, )

The difference T (N1 ) T (0) in eq. (49) can be approximated by

T
T (N1 ) T (0) N1
h h=0
and we get

1 T
N2 N1 N1
(0) h h=0

(50)

(51)

(52)

The factor on the right hand side which scales N1 is in the order of 104 or smaller, i.e. N2 N1 is in
the order of some millimetres. That means we can expect (if eq. 41 converges fast, i.e. if the step size
decays rapidly) that N1 is a good approximation of N and with eq. (50) we can define

1 T
T (N1 , , )

N1 1 +
N2 =
(0, )
(0) h h=0

(53)

which should be even better.

Usually we dont know the potential inside the masses, therefore we will do the following: We replace
Wa , the attraction part of W , by its harmonic downward continuation Wac (eq. 17) and thus T by T c
(eq. 19) and compute an associated geoid height N c (, ) which is also an approximation of the real geoid
height N . Then we try to calculate (approximately) the difference N N c caused by the masses above
the geoid. Analogous to the iterative calculation of the geoid N from the potential W by eq. (41) the
calculation of N c from W c writes:
c
Ni+1
(, ) = Nic (, ) +

1 c c
W (Ni , , ) U0
(0, )

(54)

Obviously eqs. (43) to (53) are also valid for the harmonic downward continued potential T c instead of
2 and Nc are good approximations for N and N c respectively:
T , thus N1 and N1c or even better N
2
N1c (, ) =

T c (0, , )
(0, )

c
c
c
2c (, ) = T (N1 , , ) N1c 1 + 1 T
N
(0, )
(0) h h=0

(55)

(56)

The convergence behaviour of the iterative solution in eq. (41) or (54) will not be discussed in (theoretical)
detail here. However, if we consider that the maximum relative differences between the ellipsoidal normal
potential U and the real potential W are in the order of 105 (as mentioned above) a very fast convergence
can be expected, which is confirmed by the practical calculations in section (5) (cf. figs. 6 and 7).
To estimate the difference N N c between the real geoid and the approximated geoid using the
2 and N
c from eqs. (53) and (56) and write
downward continued potential we use N
2
N (, ) N c (, )

1
1
T (N1 , , ) T c (N1c , , )
T (N, , ) T c (N, , )
(0, )
(0, )

(57)

To estimate the difference T T c in eq. (57) by using information about the topography we introduce the
potential Vt (h, , ), induced by all topographical masses above the geoid, which describes the potential
10

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

also inside the masses, and the potential Vtc (h, , ), the harmonic downward continuation of Vt . To
get the harmonic downward continuation T c of the disturbing potential T down to the geoid, we must
downward continue only the part of the potential caused by the topographic masses and write
T c (N, , ) = T (N, , ) Vt (N, , ) + Vtc (N, , )

(58)

Then for eq. (57) we get

N (, ) N c (, )

1
Vt (N, , ) Vtc (N, , )
(0, )

(59)

N (, ) N c (, ) +

1
Vt (N, , ) Vtc (N, , )
(0, )

(60)

and for the geoid:

To find an approximation of the difference Vt Vtc in eq. (59) lets treat the potential Vs (r) of a spherical
mass shell from radius R1 to R2 , with constant mass density (see fig. 3).

topography
spherical shell with
density

H
N

geoid

(,)

ellip
soid

R2
R1

Figure 3: Approximation of the topography at one point (, ) by a spherical shell of homogenous

mass density
The potential outside the shell is

GMs
for r R2
(61)
r
where Ms is the total mass of the shell and G is the gravitational constant. For r R2 it is the same
potential than that of a point mass with the same total mass Ms located at the origin of the coordinate
system. Hence the downward continuation of eq. (61) is simply the same formula defined for smaller
values of r:
GMs
for r R1
(62)
Vsc (r) =
r
The mass Ms of the shell is
4 3
Ms =
(R2 R13 )
(63)
3
Vs (r) =

11

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

Below the shell (inside the mass free inner sphere) the potential is constant:
Vs (r) = 2 G (R22 R12 )

for

r R1

(64)

Using this simple spherical shell approximation for the topographical masses with the height H above
the geoid at one specific point (, ), we get for Vtc , the downward continued potential on the geoid, from
eqs. (62) and (63) with r = R1 and R2 = R1 + H

4 G (R1 + H)3 R13

H3
c
2
Vt (N )
(65)
= 4 G R1 H + H +
3
R1
3R1
and for Vt , the potential inside the masses on the geoid, from eq. (64)

Vt (N ) 2 G (R1 + H)2 R12 = 2 G (2R1 H + H 2 )

Thus the difference Vt (N ) Vtc (N ) can be approximated by

2H 3
c
2
2
Vt (N ) Vt (N ) 2 G 2R1 H + H 2R1 H 2H
3R1

3
2H
2H
2
2
2 G H +
= 2 GH
1+
3R1
3R1

(66)

(67)

and if we neglect the second term due to H R1 (in this approximation R1 is the distance to the Earths
centre and H/R1 is in the order of 104 ):
Vt (N ) Vtc (N ) 2 GH 2

(68)

Hence we get the approximation of N N c (eq. 59) at the coordinates and to

N (, ) N c (, )

2 G H 2 (, )
(0, )

(69)

Thus, if we have a mathematical representation of the disturbing potential T c (h, , ) and if we know
the topography, i.e. the height H(, ), we can calculate an approximation N1s (superscript s stands
for spherical shell approximation) of the geoid height from eq. (55) and eq. (69) by
N (, )

N1s (, )

N1c (, )

T (0, , ) 2 G H 2 (, )
2 G H 2 (, )

=
(0, )
(0, )

(70)

c (eq. 56) instead of N c (eq. 55), by an approximation N

s:
or, with N
2
1
2
2s (, ) = N
2c (, )
N (, ) N

c c

T (N1 , , ) 2 G H 2 (, )
2 G H 2 (, )
=
(0, )
(0, )

(71)

To calculate the geoid with high accuracy the use of a more sophisticated model for the potentials
of the topography Vt and Vtc then the simple spherical shell will be inevitable. However, as seen from
eq. (59), only the difference between the potential of the topography and its downward continuation has
to be approximated instead of the potential itself; therefore the approximation in eq. (68) is more realistic
than the approximations for Vt and Vtc themselves.

12

Scientific Technical Report 09/02

3.2

Deutsches GeoForschungsZentrum GFZ

The Height Anomaly

The formulas to calculate the height anomaly from the geopotential can be derived similarly to those for
the geoid in section 3.1. With a first approximate value i for and with

U
(72)
U (ht ) U (ht i ) ( i )
h ht i
(cf. eq. 46) where ht is the ellipsoidal height of the Earths surface from eq. (12), we can write for the
difference U (ht ) U (ht i ) at point (, ) the linearisation

U
U (ht ) U (ht i ) ( i )
(73)
h ht i
By replacing U (ht ) with W (ht ) according to the definition of the height anomaly in eq. (20), and
with eqs. (47) it follows
W (ht ) U (ht i ) (ht i ) ( i )
(74)
resulting in a better approximation i+1 given by
i+1 = i +

1
W (ht ) U (ht i )
(ht i )

(75)

If we replace (ht i ) by (ht ), use W = U + T (eq. 10) and the linearisation

U (ht i ) U (ht ) + (ht )i

(76)

we again, with the start value of 0 = 0 for i = 0, get Bruns formula

1 (, ) =

T (ht , , )
(ht , )

(77)

but evaluated here for h = ht instead of h = 0 in eq. (43). Consider eq. (19), from which follows:
T c (h, , ) = T (h, , ), for h ht . The first approximation g1 of the generalised height anomaly g
for arbitrary height h as defined in eq. (25) is then:
g1 (h, , ) =

T (h, , )
(h, )

(78)

An additional approximation without using topography information, i.e. ht = 0, gives:

e1 (, ) = g1 (0, , ) = N1c (, ) =

T c (0, , )
(0, )

(79)

which is sometimes called pseudo-height anomaly calculated on the ellipsoid and is identical with
N1c (, ), the approximation of the geoid from eq. (55). Using eqs. (77) and (79), the linearisation
T (ht ) T (0) + ht

T c
h

(80)

and (0) instead of (ht ) an approximation 1 for the height anomaly on the Earths surface can also be
calculated by:

ht
T c

1 (, ) e1 (, ) +

(81)
(0) h h=0
13

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

which is useful for fast practical calculations (see section 5). In the iteration of eq. (75) we (similarly to
eqs. 40 and 41) can replace (ht i ), the normal gravity at the Telluroid, by (0), the normal gravity
at the ellipsoid, and expect that the convergence behaviour of this iteration will not change:
i+1 (, ) = i (, ) +

1
W (ht , , ) U (ht i , , )
(0)

(82)

In contrast to eq. (41) for the geoid height, where the value of the normal potential at the ellipsoid U0
is the nominal value and the iteration searches for the height h = N where the real potential has the
same value W (N ) = U0 , here, in eq. (82), for the height anomaly, the value W (ht ) for the real potential
at the Earths surface is the target value and the iteration searches for the height ht where the normal
potential has the same value U (ht ) = W (ht ). In both cases one looks for the distance of two points
lying on the same normal plumbline and the normal potential at one point must have the same value as
the real potential at the other point. But, for the geoid height N it is done near the ellipsoid, and for the
height anomaly it is done near the Earths surface. However, the main part of the difference between N
and does not come from the different ellipsoidal height where the two potentials are compared, but from
the fact, that for the geoid the real potential has to be evaluated inside the masses for points over the
continents (apart from some exotic regions) and on top (outside) of the masses for the height anomaly.
Because of the mathematical similarity of the problems one can expect, that the convergence behaviour
of the iteration in eq. (82) should be very similar to that of eq. (54) which is confirmed by the numerical
investigations in section (5).

3.3

The Difference: Geoid - Height Anomaly

Now we can estimate the difference between the geoid and the height anomaly. Considering eq. (59) and
taking N c N1c = e1 from eq. (79) the difference between the geoid and the pseudo-height anomaly (i.e.
the height anomaly on the ellipsoid) is:

1
N (, ) e (, )
Vt (N, , ) Vtc (N, , )
(83)
(0, )
With the approximation of eq. (68) for Vt Vtc we get:
N (, ) e (, )

3.4

2G H 2 (, )
(0, )

(84)

The Gravity Disturbance

If we have a mathematical representation of the potential W (h, , ), the calculation of g(h, , ) is no

problem. Using arbitrary rectangular coordinates (u, v, w) (global cartesian or local moving trihedron)
the gradients W and U of the gravity potential W and the normal potential U , i.e. the vectors of the
gravity and the normal gravity, at a given point (h, , ), are:
W (h, , ) =
U (h, , ) =

Wu (h, , ) ~eu + Wv (h, , ) ~ev + Ww (h, , ) ~ew

Uu (h, , ) ~eu + Uv (h, , ) ~ev + Uw (h, , ) ~ew

(85)
(86)

where Wu , Wv , Ww , Uu , Uv , Uw are the partial derivatives and ~eu , ~ev , ~ew are the unit vectors pointing in
the direction of u, v and w. Consequently the gravity disturbance can be calculated exactly from eq. (27):
g(h, , ) =

q
2
2
2
Wu (h, , ) + Wv (h, , ) + Ww (h, , )
q
2
2
2
Uu (h, , ) + Uv (h, , ) + Uw (h, , )

14

(87)

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

One approximation possibility is to use the fact, that the directions of the real gravity vector W and
the normal gravity vector U nearly coincide. For this purpose we write eq. (27) in the form:

W
U
g =
W

U
(88)
|W |
|U |
with the scalar product notation of eq. (36). Approximating the direction of W by U we get:

U
W U
g
|U |

(89)

and can use the disturbing potential T to write

g(h, , )

U (0, )
T (h, , )
T (h, , ) =
|U (0, )|
h

(90)

and, if h < ht , for the harmonic downward continuation g c :

g c (h, , )

T c (h, , )
h

(91)

U
The unit vector |U
| points in direction of the gradient of the normal potential, i.e. it is the normal
plumb line direction. Thus, g is also (at least approximately) the ellipsoidal normal component of the
gravity disturbance vector ~g (e.g. Hofmann-Wellenhof & Moritz, 2005).
An additional approximation is to take the direction of the radius of spherical coordinates (r, , )
instead of the ellipsoidal normal and calculate g for h = 0, i.e. on the ellipsoid, using the downward
continuation T c of the disturbing potential:

g(0, , ) gsa (0, , ) =

3.5
3.5.1

T c
r h=0

(92)

The Gravity Anomaly

The Classical Gravity Anomaly

If we have a mathematical representation of the potential W (h, , ) the calculation of gcl (, ) from
eq. (30) is no problem, however, the geoid height N has to be calculated beforehand. With eqs. (85)
and (86) we get:
gcl (, ) =

q
2
2
2
Wuc (N, , ) + Wvc (N, , ) + Wwc (N, , )
q
2
2
2
Uu (0, , ) + Uv (0, , ) + Uw (0, , )

But again an approximation of gcl in terms of the disturbing potential T is possible. With

U (0) U (N ) N |U |
h h=0

(93)

(94)

we get for eq. (30)

|U ()|
gcl (, ) W c (N, , ) U (N, ) + N

h
h=0
15

(95)

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

With eq. (28) for the downward continuation of the gravity disturbance to h = N and eq. (16) for the
normal gravity we can write it in the more usual form

()
c
gcl (, ) g (N, , ) + N
(96)
h h=0
With the approximations (91) for g c and (56) for N we get

T c (N1c , , ) ()
T c (h, , )
+
gcl (, )

h
(0, )
h h=0
h=N

(97)

to calculate the classical gravity anomaly from the disturbing potential. Without previous knowledge of
N , i.e. using eq. (55) instead of (56), we can approximate the classical gravity anomaly to:
gcl (, )

T c (0, , ) ()
T c (h, , )
+

h
(0, )
h h=0
h=0

(98)

If we again replace the ellipsoidal normal by the radial direction and approximate additionally the normal
gravity by its spherical term
()

GM
()
2 GM
and
3
2
r ()
r
r ()

(99)

(G is the gravitational constant and M is the mass of the Earth) we get the spherical approximation
gsa of the classical gravity anomaly to:

T c
2
gcl (, ) gsa (, ) =
T c (0, , )

r h=0
r()

(100)

where r = r() is the distance to the centre of the coordinate system (spherical coordinate) of a point on
the ellipsoid.
3.5.2

The Modern Gravity Anomaly

The calculation of g using eq. (31) at a given point (h, , ) is possible if the height anomaly has been
calculated beforehand:
q
2
2
2
g(h, , ) =
Wu (h, , ) + Wv (h, , ) + Ww (h, , )
q
2
2
2
Uu (h , , ) + Uv (h , , ) + Uw (h , , )

(101)

To calculate it in terms of the disturbing potential analogously to the classical gravity anomaly we get:

()
, for h ht
(102)
g(h, , ) g(h, , ) +
h h=0
Alike eq. (98) the approximation without knowledge of (i.e. = 0) is:
g(h, , )

T c (h, , )
T c (0, , ) ()

h
(0, )
h h
h
16

(103)

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

which is valid for arbitrary points outside the geoid because T c has been used (remember that T c = T
for h ht ). For h = 0 eq. (103) is the same as eq. (98) for the classical gravity anomaly gcl , so the
spherical approximation using eq. (99) is the same too:

T c
2
g(, ) gsa (, ) =

T c (0, , )
r h=0
r()
3.5.3

(104)

The Topography-Reduced Gravity Anomaly

If we know the density (h, , ) of the masses above the geoid and the height H(, ) of the topography
above the geoid the calculation of the potential Vt (h, , ) (and its derivatives) in eqs. (33) and (34) is in
principle possible by numerical integration. But, however, it is extensive. Therefore, in the past, without
todays computer power, the question was: how can the potential Vt of the topographical masses be
replaced by a first approximation which results in a simple formula depending only on a constant density
and the height H(, ) of the point where it should be calculated?
As a simple but useful approximation for the topography-reduced gravity anomaly, the Bouguer
anomaly gB has been introduced as:

gt r(, ) gB (, ) = W c (N, , ) AB U (0, )

(105)
where
AB (, ) = 2GH(, )

(106)

is the attraction of the so called Bouguer plate, which is a plate of thickness H (topographical height
above geoid), constant density and infinite horizontal extent (e.g. Hofmann-Wellenhof & Moritz, 2005).
With the classical gravity anomaly (eq. 30) we get:
gB (, ) = gcl (, ) 2GH(, )

(107)

Unfortunately this cannot be expressed in terms of potentials because the potential of an infinite plate
makes no sense (cf. the discussion in Vancek et al., 2001 and Vancek et al., 2004). The obvious idea
to use the potential of a spherical shell as in section 3.1 results in a contribution of 4GH which is
twice the Bouguer plate-attraction. The plausible explanation is that the contribution of the far zone
of the spherical shell, the whole opposite half sphere, cannot be neglected here, whereas in eq. (59), for
the difference Vt (N ) Vtc (N ) (which results in eq. 69), it can.
To find a simple approximation for Vt which is consistent with the results for the Bouguer plate one
could define the potential of a spherical cap of constant thickness H, or a gaussian bell shaped mountain
with height H, and an extend which produce the attraction of AB = 2GH.

4
4.1

Calculation from Spherical Harmonics

Spherical Harmonics and the Gravity Field

The solid spherical harmonics are an orthogonal set of solutions of the Laplace equation represented
in a system of spherical coordinates. (e.g. Hobson, 1931; Freeden, 1985; Hofmann-Wellenhof & Moritz,
2005). Thus, each harmonic potential, i.e. such which fulfils Laplaces equation, can be expanded into
solid spherical harmonics. For this reason the stationary part of the Earths gravitational potential Wa
(the attraction part only, see eq. 5) at any point (r, , ) on and above the Earths surface is expressed
on a global scale conveniently by summing up over degree and order of a spherical harmonic expansion.
The spherical harmonic (or Stokes) coefficients represent in the spectral domain the global structure and

17

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

irregularities of the geopotential field or, speaking more generally, of the gravity field of the Earth. The
equation relating the spatial and spectral domains of the geopotential is as follows:
Wa (r, , ) =

`max X
` `
W

GM X
R
W
P`m (sin ) C`m
cos m + S`m
sin m
r
r
m=0

(108)

`=0

which shows the 1/r-behaviour for r , or written in the form

Wa (r, , ) =

`max X
` `+1

W
GM X
R
W
cos m + S`m
sin m
P`m (sin ) C`m
R
r
m=0
`=0

which is sometimes useful in practice. The notations are:

r, ,

R
GM
`, m
P`m
W
W
C`m
, S`m

spherical geocentric coordinates of computation point

(radius, latitude, longitude)
reference radius
product of gravitational constant and mass of the Earth
degree, order of spherical harmonic
fully normalised Lengendre functions
Stokes coefficients (fully normalised)

The transformation formulas between ellipsoidal (h, , ) and spherical (r, , ) coordinates can be found
e.g. in (Hofmann-Wellenhof & Moritz, 2005). A spherical harmonic approximation of the gravity field
up to a maximum degree `max (a so-called gravity field model) consists of (`max + 1)2 coefficients and
the 2 values for GM and R to which the coefficients relate. The reference radius R of the expansion
W
GM represents the
has only mathematical meaning. As can be seen from eq. (108), the product C00
W
gravitational constant times the mass of the Earth associated with the model. This means that C00
W
scales the formal value of GM which is given with the model. Usually C00 is defined to 1 to preserve
the meaning of GM which itself is not separated into its two single values G (gravitational constant)
and M (mass of the Earth) because it is known as product with a much higher accuracy than the two
W
W
W
) are related to the geocentre
, S11
, C11
separate values. The degree 1 spherical harmonic coefficients (C10
coordinates and are zero if the coordinate systems origin coincides with the geocentre. The coefficients
W
W
are connected to the mean rotational pole position.
and S21
C21
Thus, eq. (108) represents the Earths gravity field with an accuracy depending on the accuracy of
W
W
the coefficients (C`m
, S`m
) and a spatial resolution depending on the maximum degree `max . At a given
point in space the difference of the real potential and the potential represented by the spherical harmonic
expansion in eq. (108) depends on both, the coefficients accuracy and the maximum degree `max of the
expansion.
Equation (108) contains the upward-continuation of the gravitational potential from the Earths surface
for r > rtopo and reflects the attenuation of the signal with altitude through the factor (R/r)` . For points
lying inside the Earth the spherical harmonic expansion gives the harmonic downward continuation Wac
of the potential in a natural way simply by evaluating it for r < rtopo . However, possible singularities
of this downward continuation (see the remarks in subsection 2.1) would result in divergence of the
spherical harmonic series at the singular points for `max (cf. the discussion of this topic for the
Earth surface in Moritz, 1989). In practise `max is finite and the series can be evaluated, in principle, also
for points lying inside the Earth (r < rtopo ). However, the harmonic downward continuation, from its
physical nature, is an unstable and ill-posed problem. That means the amplitudes of spatial undulations
of the potential are amplified with depth (up to infinity at the locations of the singularities) and the
amplification is bigger the shorter the wavelength of the undulation is. Mathematically this is obvious
from the factor (R/r)` in eq. (108) for decreasing radius r and increasing degree `. Thus, downward
continuation in practise is always a (frequency-dependent) amplification of errors, i.e. in case of spherical
W
W
harmonic representation an `-dependent amplification of the errors of the coefficients C`m
, S`m
.
18

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

Note, that the spherical harmonics are calculated using spherical coordinates, so rtopo = rtopo (, )
is the distance of the point on the topography (Earths surface) from the Earths centre and is the
spherical latitude to be distinguished from the ellipsoidal latitude .
Figure 4 presents examples for the three different kinds of spherical harmonics P`m (sin ) cos m: (a)
zonal with l 6= 0, m = 0, (b) tesseral with l 6= 0, m 6= l 6= 0 and (c) sectorial harmonics with ` = m 6= 0.

zonal: ` = 6, m = 0

tesseral: ` = 16, m = 9

sectorial: ` = 9, m = 9

Figure 4: Examples for spherical harmonics P`m (sin ) cos m [from 1 (blue) to +1 (violet)]
Obviously, the attraction part Ua of the normal potential U (see eq. 13), and thus the disturbing
potential T , according to eq. (10), can be expanded into spherical harmonics too. If we denote the
U
U
T
T
coefficients which represent Ua by C`m
, S`m
the coefficients C`m
, S`m
of the disturbing potential are
simply the differences
T
W
U
T
W
U
C`m
= C`m
C`m
and
S`m
= S`m
S`m
(109)
The expansion of the ellipsoidal normal potential contains only terms for order m = 0 (rotational symT
W
metry) and degree ` = even (equatorial symmetry). Recall that S`,0 dont exist, so S`,m
= S`,m
. To
calculate the normal potential in practise in most cases it is sufficient to consider only the coefficients
U
U
U
U
U
. The disturbing potential T in spherical harmonics is:
and sometimes C80
, C60
, C40
, C20
C00
`max ` X
`
T

GM X
R
T
T (r, , ) =
P`m (sin ) C`m
cos m + S`m
sin m
r
r m=0

(110)

`=0

for r rtopo . For r < rtopo equation (110) gives T c , the harmonic downward continuation of T , introduced
in section (3.1).
U
Here, we implicitly postulated that C`m
, the coefficients of the attraction part of the normal potential,
W
W
and C`m and S`m , the coefficients of the real potential (or the potential of a model approximating the
real potential), are given with respect to the same values for GM and R. Usually this is not the case in
U
practise where the normal potential coefficients C`m
are given with respect to separately defined values
U
U
GM and R . From comparing the summands of the series separately, which must be equal due to
orthogonality, the relation between them is found to be:

`
GM U RU
U
U
(111)
C`m
= C`m

GM
R
and must be considered in eq. (109).
Each representation of a function in spherical harmonics like eq. (108) with an upper limit of summation
`max < corresponds to a low pass filtering, and `max correlates to the spatial resolution at the Earth
surface. A usual simple estimation of the smallest representable feature of the gravity field, in other
words, the shortest half-wavelength min (as spherical distance), that can be resolved by the (`max + 1)2
parameters C`m , S`m is:
R
(112)
min (`max )
`max
19

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

This estimation is based on the number of possible zeros along the equator.
At this point let us recall that the resolution of spherical harmonics is uniform on the sphere. This
follows from the known fact that under rotation, a spherical harmonic of degree ` is transformed into
a linear combination of spherical harmonics of the same degree. To illustrate it, imagine a single pulse
somewhere on the sphere represented (as narrow as possible) by spherical harmonics up to a maximum
degree and order `max . A rotation of the coordinate system will not change the shape of the pulse which
means uniform resolution. Hence, a better estimation of min (`max ) seems to be the following: If we
divide the surface of the sphere, i.e. 4R2 , into as many equiareal pieces Amin as the number of spherical
harmonic coefficients, i.e. (`max + 1)2 , then the size of each piece is:
Amin (`max ) =

4R2
(`max + 1)2

(113)

The diameter of a spherical cap of this size is (in units of spherical distance):

1
min (`max ) = 4 arcsin
`max + 1

(114)

which characterise the size of the smallest bump, half-wavelength, which can be produced by (`max + 1)2
parameters. For some selected maximum degrees the resolutions are given in Table 1. To demonstrate
Table 1: Examples of spatial resolution of spherical harmonics in terms of the diameter min of the
smallest representable shape (bump or hollow) after eqs. (112) and (114)
Maximum

Number of

Degree

Coefficients

`max

2
5
10
15
30
36
40
45
50
75
180
360

9
36
121
256
961
1369
1681
2116
2601
5776
32761
130321

Resolution min
eq. (112)
[degree]
90.0
36.0
18.0
12.0
6.0
5.0
4.5
4.0
3.6
2.4
1.0
0.5

eq. (114)

[km]

[degree]

[km]

10000.000
4000.000
2000.000
1333.333
666.667
555.556
500.000
444.444
400.000
266.667
111.111
55.556

77.885
38.376
20.864
14.333
7.394
6.195
5.590
4.983
4.494
3.016
1.266
0.635

8653.876
4264.030
2318.182
1592.587
821.587
688.321
621.154
553.626
499.342
335.073
140.690
70.540

how the resolution of spherical harmonics depends on the maximum degree `max of the development the
following synthetic example has been constructed: A (1 1 )-grid where all elements are zero except
for two with the values 1 has been converted into spherical harmonic coefficients up to degree and order
`max = 90 using the numerical integration described in (Sneeuw, 1994). The two peaks are 6 (spherical
distance) apart from each other. The cross-sections through the two peaks for different maximum degrees
`max are shown in Figure (5). From Table (1) (eq. 114) one expects that a maximum degree of `max 36
suffices to resolve the peaks but the result of this example is that a slightly higher maximum degree of
`max 41 is necessary.
20

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

Figure 5: Cross-sections through 2 peaks, which are originally 6 apart, after approximation by
spherical harmonics of different maximum degrees `max

4.2

The Geoid

To calculate the geoid undulation from eq. (70) or eq. (71) besides the potentials W and U , or T = W U ,
a representation of the topography H(, ) must be available too. Usually the topography models are
given as grids on the reference ellipsoid and have a much higher resolution (e.g. 20 20 ) than the
recent global gravity field models. To avoid adding parts of different resolution in eqs. (70) or (71), the
topography model can also be transformed into a surface spherical harmonics expansion (a good summary
for this technique can be found in Sneeuw, 1994):
H(, ) = R

`X
`
max X

topo

topo
P`m (sin ) C`m
cos m + S`m
sin m

(115)

`=0 m=0
topo
topo
where C`m
and S`m
are the coefficients of the expansion which are usually scaled by the reference
radius R. Using the same upper limit `max of the expansion, the geoid N can be approximated according

21

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

to eq. (70) by:

N1s (, ) =

` `
`X
max
T

R X
GM
T
P`m (sin ) C`m
cos m + S`m
sin m
re (re , )
re m=0
`=0

"
2 G R

`X
`
max X

topo
topo
P`m (sin ) C`m
cos m + S`m
sin m

#2

(116)

`=0 m=0

or according to eq. (71) by:

2s (, ) =
N

` `
`X
max
T

R X
GM
T
P`m (sin ) C`m
cos m + S`m
sin m
r (re , )
r m=0
`=0

"
2 G R

`X
`
max X

topo

topo
P`m (sin ) C`m
cos m + S`m
sin m

#2

(117)

`=0 m=0

The radius-coordinate of the calculation point for the normal gravity is set to the latitude dependent
radius-coordinate re = re () of points on the ellipsoid, and for the disturbing potential it is set to
r = re () as well if eq. (70) is used, or to r = r (, ) if eq. (71) is used, whereas in the latter case an
approximation for e = e (, ) has to be calculated in a prior step using eq. (118).

4.3

The Height Anomaly

The calculation of the height anomalies (, ) from spherical harmonic potential models according to
the iteration (eq. 82) is possible using eqs. (4), (5) and (108). Using the coefficients of the disturbing
potential (eqs. 109, 110 and 111) the calculation according to eq. (77) or (79) is simple. Without using
a topography model they can be calculated from eq. (79) by
` `
`X
max
T

GM
R X
T
P`m (sin ) C`m
cos m + S`m
sin m
e1 (, ) =
re (re , )
re m=0

(118)

`=0

and the radius-coordinate r(, ) of the calculation point is be set to re = re (). With eq. (81) they can
be calculated more accurately by

H(, ) + N (, ) T c

1 = e1 +

(0, )
r r=re

(119)

which can be calculated from spherical harmonics if we use eq. (115) for H and eq. (116) or (117) for N ,
and eqs. (92) and (125) for the radial derivative of T c .

4.4

The Gravity Disturbance

To calculate the gravity disturbance from eq. (87) the gradients W from eq. (85) and U from eq. (86)
have to be calculated from spherical harmonics. The gradient W in spherical coordinates is (e.g. Bronshtein et al., 2004):
1
1
W~e + W~e
(120)
W = Wr ~er +
r cos
r
22

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

where Wr , W , W are the partial derivatives and ~er , ~e , ~e are the unit vectors pointing in the direction
of r, and respectively. Consequently for |W |, considering the centrifugal potential (eqs. 4 and
5), we have:
s

2
2

1
1
W = [War + r ]2 +
(121)
(Wa + ) + (Wa + )
r cos
r
The derivatives of eq. (108) in spherical harmonics are:
War

`max `
`
X
W

GM X
R
W
= 2
(` + 1)
P`m (sin ) C`m
cos m + S`m
sin m
r
r
m=0
`=0

GM
r

Wa =

`X
max
`=0

R
r

` X
`

W
sin m
m P`m (sin ) S`m cos m C`m

(122)

m=0

`max ` X
`

GM X
R
P`m (sin) W
W
C`m cos m + S`m
sin m
r
r m=0

Wa =

`=0

The centrifugal potential (eq. 4) in spherical coordinates is

=

1 2 2
r (cos )2
2

(123)

and its derivatives are:

r = 2 r (cos )2

= 0

= 2 r2 cos sin

(124)

hence |W | can be calculated from eq. (121) with eqs. (122) and (124).
W
W
U
U
The same formulas hold for calculating |U | by replacing W by U and C`m
, S`m
by C`m
, S`m
, thus the
gravity disturbance g outside the masses or its downward continuation g c can be calculated exactly for
arbitrary points (r, , ) in space from the spherical harmonic coefficients of a given gravity field model.
The spherical approximation gsa of the gravity disturbance (eq. 92), i.e. the radial derivative of the
disturbing potential T , calculated from a spherical harmonic expansion of T is:
`max `
`
X

T
GM X
R
T
sin m
gsa (r, , ) = 2
(` + 1)
P`m (sin ) C`m
cos m + S`m
r
r
m=0

(125)

`=0

4.5

The Gravity Anomaly

Similar to the gravity disturbance g, equations (121) to (124) can be used to calculate the gravity
anomaly g exactly from spherical harmonics from eq. (101) (or from eq. (93) for the classical gcl ).
The spherical approximation gsa (eq. 104) calculated from a spherical harmonic expansion of the
disturbing potential T is:
gsa (r, , ) =

`max `
`
X
T

GM X
R
T
(`

1)
P`m (sin ) C`m
cos m + S`m
sin m
2
r
r
m=0

(126)

`=0

Nowadays this formula, as well as the spherical approximation of the gravity disturbance (eq. 125), is
not accurate enough for most practical applications. Nevertheless the degree-dependent factors (` + 1 for
gsa and ` 1 for gsa ) give theoretical insight into the different spectral behaviour and are useful for
simulation studies.
23

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

Practical calculations using the model EIGEN-GL04C

The recent gravity field model EIGEN-GL04C (Forste et al., 2006) and the topography model ETOPO2
(U.S. Department of Commerce, 2001) have been used to calculate the described functionals and their
approximations on global grids. All calculations are carried out with respect to the reference system
WGS84. The differences between several approximations of one and the same functional are presented.
The topography model is necessary for two different purposes: (a) to calculate the exact coordinates
on the Earths surface for the height anomalies on the Earths surface, the gravity disturbances and the
modern gravity anomalies, and (b) to calculate the geoid undulations from pseudo height anomalies on
the ellipsoid considering the topographical effect. For (a) bi-linear interpolation of the original ETOPO2grid is used to calculate the positions as accurately as possible. For (b) the gridded values of the original
topography model has been transformed into a (surface-) spherical harmonic expansion using the formulas
described by Sneeuw (1994). The maximum degree of EIGEN-GL04C is `max = 360, so the topography
has been used with the same resolution.

5.1

Geoid and Height Anomaly

The convergence of the iterative calculation of the downward continued geoid N c (, ) from W c (, )
and U0 using eq. (54) with N1c from eq. (55) or (79) as start value is very fast. For the calculation using
spherical harmonics eqs. (108) and (118) are used for W c and N1c respectively. With a 64-bit accuracy
(real*8 in Fortran, which are about 16 decimal digits) convergence is reached after 2 or 3 steps (step 1
is from N1c to N2c in our terminology). The spatial distribution of the accuracy of the start value N1c
from eq. (55) and the first iteration N2c of eq. (54) are demonstrated by figures (6) and (7). Figure (8)
c of eq. (56). Thus, to calculate the real geoid
shows the accuracy of the (non-iterative) approximation N
2
c
N from the downward continued geoid N by adding the influence of the topography (see eq. 60) the
calculation of N c has to be carried out with the required accuracy. However, as mentioned in section (3.1),
the influence of the topographic masses are the accuracy limiting factor.
If we start the iterative calculation of the height anomalies (ht , , ) (eq. 82) with the pseudo height
anomaly on the ellipsoid e1 according to eqs. (79) and (118) the mean difference between start values and
convergence is about 30 times bigger than for the iterative calculation of N c using eq. (54). Nevertheless
the iteration converges after 3 or 4 steps. The differences between the start values e1 and the convergence
are shown in figure (9) and the accuracy of the first iteration 1 is shown in figure (10). Figure (11) shows
the accuracy of the (non-iterative) approximation 1 of eqs. (81) and (119). Usually the chosen calculation
formula or algorithm will be a compromise between computing effort and accuracy requirement.
Figure (12) shows the differences between the height anomalies on the ellipsoid from eqs. (79) and (118)
and the geoid heights from eqs. (71) and (117) (calculated using the spherical shell approximation for the
topography). The maximum exceeds 3 metres. Thus, to calculate geoid undulations in continental areas,
the potential of the topographic masses must be approximated somehow at the geoid inside the masses.
The difference between the potential of the topographic masses and its harmonic downward continuation,
both evaluated on the geoid, cannot be neglected.
The weighted means, the minima and the maxima of the grid differences shown in figures (6) to (12)
are summarised in table (2).

5.2

Gravity Disturbance and Gravity Anomaly

Figure 13 shows the differences (gcl g) between the classically defined and the modern (Molodenskys
theory) gravity anomalies. They range from 12 to +24 mgal. Therefore, when using gravity anomaly
data sets which are derived from real measurements it is important to know how they are reduced.
The differences (gsa g) between spherically approximated and exactly calculated gravity disturbances are nearly the same as (gsa g), the differences between spherically approximated and exact
modern (Molodenskys theory) gravity anomalies. The differences (gsa g) are shown in figure 14.

24

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

Table 2: Differences between varying approximation levels of height anomalies and geoid undulations
Differences
[m]

wrms
[m]

min
[m]

max
[m]

N1c N c
N2c N c
c Nc
N
2
e1
2
1
s
e1 N
2

1.3 103
1.4 107
2.4 104
3.8 102
4.7 105
6.5 103
2.4 101

1.5 102
5.7 106
1.8 103
3.5 101
3.0 104
1.1 101
0.0

2.2 102
3.8 106
5.0 105
1.1
2.0 103
8.5 102
3.67

They range from 13 to +23 mgal. To compare or combine gravity anomalies or gravity disturbances
derived from terrestrial measurements with those derived from gravity field models in spherical harmonic
representation the spherical approximation is obviously not accurate enough.
The differences (gsa gcl ) between spherically approximated and classical gravity anomalies are
smaller, they range from 1.48 to +0.83 mgal and are shown in figure 15. Hence, to calculate classical
gravity anomalies the spherical approximation can be used if the highest accuracy is not required.
The weighted means, the minima and the maxima of the grid differences shown in figures (13) to (15)
are summarised in table (3).
Table 3: Differences between the gravity anomalies, the classical gravity anomalies and the gravity
anomalies in spherical approximation
Differences
[mgal]

wrms
[mgal]

min
[mgal]

max
[mgal]

gcl g
gsa g
gsa gcl

0.73
0.69
0.11

12
13
1.48

26
23
0.83

25

Scientific Technical Report 09/02

0.005

0.004

0.003

Deutsches GeoForschungsZentrum GFZ

0.002

0.001

0.000

0.001

0.002

0.003

0.004

0.005

Figure 6: The difference between the start value N1c and the convergence N c of eq. (54):
(N1c N c ); wrms = 1.3 103 m, min = 0.015 m, max = 0.022 m

5e07

4e07

3e07

2e07

1e07

0e+00

1e07

2e07

3e07

4e07

5e07

Figure 7: The difference between the first iteration N2c and the convergence of eq. (54):
(N2c N c ); wrms = 1.4 107 m, min = 5.7 106 m, max = 3.8 106 m

26

Scientific Technical Report 09/02

1e03

8e04

6e04

Deutsches GeoForschungsZentrum GFZ

4e04

2e04

3e20

2e04

4e04

6e04

8e04

1e03

Figure 8: The difference between the approximation in eq. (56) and the convergence of eq. (54):
(Nc N c ); wrms = 2.4 104 m, min = 1.8 103 m, max = 5 105 m
2

0.20

0.16

0.12

0.08

0.04

0.00

0.04

0.08

0.12

0.16

0.20

Figure 9: The difference between the pseudo height anomalies on the ellipsoid and the height
anomalies on the Earths surface: (e1 ); wrms = 0.038 m, min = 0.35 m,
max = 1.1 m

27

Scientific Technical Report 09/02

5e04

4e04

3e04

Deutsches GeoForschungsZentrum GFZ

2e04

1e04

1e20

1e04

2e04

3e04

4e04

5e04

Figure 10: The differences between the first iteration 2 and the convergence of eq. (82):
(2 ); wrms = 4.7 105 m, min = 3.0 104 m, max = 2.0 103 m

0.040

0.032

0.024

0.016

0.008

0.000

0.008

0.016

0.024

0.032

0.040

Figure 11: The difference between the approximation 1 of eq. (81) and the convergence of
eq. (82): (1 ); wrms = 6.5 103 m, min = 0.11 m, max = 0.085 m

28

Scientific Technical Report 09/02

1.0

0.8

0.6

Deutsches GeoForschungsZentrum GFZ

0.4

0.2

0.0

0.2

0.4

0.6

0.8

1.0

Figure 12: The difference between pseudo height anomalies on the ellipsoid of eq. (118) and
s ); wrms = 0.24 m, min = 0.0 m,
geoid undulations from eq. (117): (e1 N
2
max = 3.67 m

3.0

2.4

1.8

1.2

0.6

0.0

0.6

1.2

1.8

2.4

3.0

Figure 13: The difference between classical and modern (Molodenskys theory) gravity
anomalies: (gcl g); wrms = 0.73 mgal, min = 12 mgal, max = 26 mgal

29

Scientific Technical Report 09/02

3.0

2.4

1.8

Deutsches GeoForschungsZentrum GFZ

1.2

0.6

0.0

0.6

1.2

1.8

2.4

3.0

Figure 14: The difference (gsa g) between spherical approximated gravity anomalies and
modern (Molodenskys theory) gravity anomalies; wrms = 0.69 mgal,
min = 13 mgal, max = 23 mgal

0.40

0.32

0.24

0.16

0.08

0.00

0.08

0.16

0.24

0.32

0.40

Figure 15: The difference (gsa gcl ) between spherical approximated and classical gravity
anomalies; wrms = 0.11 mgal, min = 1.48 mgal, max = 0.83 mgal

30

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

References
Blakely, R., 1995, Potential theory in gravity and magnetic applications, Cambridge Univ. Press, Cambridge.
Bronshtein, I. N., Semendyayev, K. A., & Musiol, G., 2004, Handbook of Mathematics, Springer, Berlin
[et al.].
Forste, C., Flechtner, F., Schmidt, R., andUlrich Meyer, R. K., Stubenvoll, R., Rothacher, M., Neumayer,
F. B. H., Biancale, R., Bruinsma, S., Jean-MichelLemoine, & Loyer, S., 2006, A mean global gravity
field model from the combination of satellite missionand altimetry/gravimetry surface data - EIGENGL04C, Geophys. Res. Abstr., 8, 03462.
Freeden, W., 1985, Computation of spherical harmonics and approximation by spherical harmonicexpansions, Report / Department of Geodetic Science and Surveying, the Ohio State University; 362,
Ohio.
Hackney, R. & Featherstone, W., 2003, Geodetic versus geophysical perspectives of the gravity anomaly,
Geophys. Jour. Intern., 154(1), 3543.
Heiskanen, W. A. & Moritz, H., 1967, Physical geodesy, A series of books in geology, Freeman, San
Francisco [et al.].
Hobson, E. W., 1931, The Theory of Spherical and Ellipsoidal Harmonics, Cambridge University Press,
Cambridge.
Hofmann-Wellenhof, B. & Moritz, H., 2005, Physical geodesy, Wien [et al.]: Springer, 2005.
Jekeli, C., 2000, Heights, the geopotential, and vertical datums, Tech. Rep. 459, Department of Civil
and Environmental Engineering and Geodedic Science,The Ohio State University, Columbus, Ohio
43210-1275.
Martinec, Z., 1998, Boundary-value problems for gravimetric determination of a precise geoid , vol. 73 of
Lecture Notes in Earth Sciences, Springer-Verlag Berlin Heidelberg New York.
Molodensky, M., Eremeev, V., & Yurkina, M., 1962, Methods for study of the external gravitational field
and figure of theEarth, Transl. from Russian (1960), Israel Program for Scientific Translations.
Moritz, H., 1980, Geodetic Reference System 1980, Bull. Geod., 54, 395405.
Moritz, H., 1989, Advanced physical geodesy, Karlsruhe: Wichmann, 1989.
NIMA, 2000, Department of defense world geodetic system 1984, Tech. Rep. NIMA TR8350.2, Third
Edition, January 2000.
Pick, M., Picha, J., & Vyskocil, V., 1973, Theory of the earths gravity field , Academia, Praha.
Sneeuw, N., 1994, Global spherical harmonic analysis by least-squares and numerical quadraturemethods
in historical perspective, Geophys. Jour. Intern., 118, 707716.
Torge, W., 1991, Geodesy, Berlin; New York: de Gruyter, 1991.
U.S. Department of Commerce, 2001, 2-minute Gridded Global Relief Data (ETOPO2),
http://www.ngdc.noaa.gov/mgg/fliers/01mgg04.html.
Vancek, P. & Christou, N. T., eds., 1994, Geoid and its geophysical interpretation, CRC Press, Boca
Raton [et al.].
31

Scientific Technical Report 09/02

Deutsches GeoForschungsZentrum GFZ

Vancek, P. & Krakiwsky, E. J., 1982, Geodesy: The Concepts, North-Holland, Amsterdam [et al.].
Vancek, P., Novak, P., & Martinec, Z., 2001, Geoid, topography, and the Boguer plate or shell, Jour.
Geod., 75(4), 210215.
Vancek, P., Tenzer, R., Sjoberg, L. E., Martinec, Z., & Featherstone, W. E., 2004, New views of the
spherical Bouguer gravity anomaly, Geophys. Jour. Intern., 159, 460472.
Zidarov, D., 1990, Inverse gravimetric problem in geoprospecting and geodesy, Developments in solid earth
geophysics ; 19, Elsevier, Amsterdam [u.a.].

32