GEOPHYSICS, VOL. 66, NO. 6 (NOVEMBER-DECEMBER 2001); P. 1660–1668, 4 FIGS., 3 TABLES.

Tutorial

Ellipsoid, geoid, gravity, geodesy, and geophysics

Xiong Li∗ and Hans-Jürgen Götze‡

ABSTRACT surement we could make accurately (i.e., by leveling).

Geophysics uses gravity to learn about the den- The GPS delivers a measurement of height above the
sity variations of the Earth’s interior, whereas classical ellipsoid. In principle, in the geophysical use of gravity,
geodesy uses gravity to define the geoid. This difference the ellipsoid height rather than the elevation should be
in purpose has led to some confusion among geophysi- used throughout because a combination of the latitude
cists, and this tutorial attempts to clarify two points of correction estimated by the International Gravity For-
the confusion. First, it is well known now that gravity mula and the height correction is designed to remove
anomalies after the “free-air” correction are still located the gravity effects due to an ellipsoid of revolution. In
at their original positions. However, the “free-air” re- practice, for minerals and petroleum exploration, use of
duction was thought historically to relocate gravity from the elevation rather than the ellipsoid height hardly in-
its observation position to the geoid (mean sea level). troduces significant errors across the region of investi-
Such an understanding is a geodetic fiction, invalid and gation because the geoid is very smooth. Furthermore,
unacceptable in geophysics. Second, in gravity correc- the gravity effects due to an ellipsoid actually can be
tions and gravity anomalies, the elevation has been used calculated by a closed-form expression. However, its ap-
routinely. The main reason is that, before the emergence proximation, by the International Gravity Formula and
and widespread use of the Global Positioning System the height correction including the second-order terms,
(GPS), height above the geoid was the only height mea- is typically accurate enough worldwide.

INTRODUCTION tions require three different surfaces to be clearly defined. They

are (Figure 1): the highly irregular topographic surface (the
Geophysics has traditionally borrowed concepts of gravity landmass topography as well as the ocean bathymetry), a geo-
corrections and gravity anomalies from geodesy. Their uncrit- metric or mathematical reference surface called the ellipsoid,
ical use has sometimes had unfortunate results. For example, and the geoid, the equipotential surface that mean sea level
the “free-air” reduction was historically interpreted by geode- follows.
sists as reducing gravity from topographic surface to the geoid Gravity is closely associated with these three surfaces. Grav-
(mean sea level). This interpretation is a useful fiction for ity corrections and gravity anomalies have been traditionally
geodetic purposes, but is completely inappropriate for geo- defined with respect to the elevation. Before the emergence
physics. In geophysics, gravity is used to learn about the den- of satellite technologies and, in particular, the widespread
sity variations of the Earth’s interior. In geodesy, gravity helps use of the Global Positioning System (GPS), height above
define the figure of the Earth, the geoid. This difference in pur- the geoid (i.e., the elevation) was the only height measure-
pose determines a difference in the way to correct observed ment we could make accurately, namely by leveling. The
data and to understand resulting anomalies. GPS delivers a measurement of height above the ellipsoid.
Until a global geodetic datum is fully and formally accepted, Confusion seems to have arisen over which height to use in
used, and implemented worldwide, global geodetic applica- geophysics.

Published on Geophysics Online May 31, 2001. Manuscript received by the Editor August 9, 2000; Revised manuscript received February 26, 2001.
∗
Fugro-LCT Inc., 6100 Hillcroft, 5th Floor, Houston, Texas 77081. E-mail: [email protected].
‡Freie Universität Berlin, Institut für Geologie, Geophysik and Geoinformatik, Malteserstraße 74-100, D-12249 Berlin, Germany. E-mail: hajo@
geophysik.fu-berlin.de.
°c 2001 Society of Exploration Geophysicists. All rights reserved.

1660
 Correctly Understanding Gravity 1661

This tutorial explains the concepts of, and relationships and World Geodetic System 1984 (WGS 84) all are “normal
among, the ellipsoid, geoid, gravity, geodesy, and geophysics. Earth.”
We attempt to clarify the way to best compute gravity correc- Although the Earth is not an exact ellipsoid, the equipoten-
tions given GPS positioning. In short, h, the ellipsoid height tial ellipsoid furnishes a simple, consistent and uniform refer-
relative to the ellipsoid, is the sum of H , the elevation relative ence system for all purposes of geodesy as well as geophysics:
to the geoid, and N , the geoid height (undulation) relative to a reference surface for geometric use such as map projec-
the ellipsoid (Figure 2): tions and satellite navigation, and a normal gravity field on
the Earth’s surface and in space, defined in terms of closed
h = H + N. (1)
formulas, as a reference for gravimetry and satellite geodesy.
The geoid undulations, gravity anomalies, and gravity gradient The gravity field of an ellipsoid is of fundamental practical im-
changes all reflect, but are different measures of, the density portance because it is easy to handle mathematically, and the
variations of the Earth. The difference between the geophys- deviations of the actual gravity field from the ellipsoidal “theo-
ical use of gravity and the geodetic use of gravity mirrors the retical” or “normal” field are small. This splitting of the Earth’s
difference between the ellipsoid and the geoid. gravity field into a “normal” and a remaining small “disturb-
ing” or “anomalous” field considerably simplifies many prob-
ELLIPSOID lems: the determination of the geoid (for geodesists), and the
use of gravity anomalies to understand the Earth’s interior (for
As a first approximation, the Earth is a rotating sphere. As a geophysicists).
second approximation, it can be regarded as an equipotential Although an ellipsoid has many geometric and physical pa-
ellipsoid of revolution. rameters, it can be fully defined by any four independent pa-
According to Moritz (1980), the theory of the equipotential rameters. All the other parameters can be derived from the
ellipsoid was first given by P. Pizzetti in 1894. It was further four defining parameters. Table 1 lists the two geometric pa-
elaborated by C. Somigliana in 1929. This theory served as rameters of several representative ellipsoids. Notice how the
the basis for the International Gravity Formula adopted at the parameters differ, depending on the choice of ellipsoid.
General Assembly of the International Union of Geodesy and One of the principal purposes of a world geodetic system is
Geophysics (IUGG) in Stockholm in 1930. One particular el- to supersede the local horizontal geodetic datums developed
lipsoid of revolution, also called the “normal Earth,” is the one to satisfy mapping and navigation requirements for specific re-
having the same angular velocity and the same mass as the ac- gions of the Earth. A particular reference ellipsoid was used to
tual Earth, the potential U0 on the ellipsoid surface equal to the help define a local datum. For example, the Australian National
potential W0 on the geoid, and the center coincident with the ellipsoid (Table 1) was used to define the Australian Geodetic
center of mass of the Earth. The Geodetic Reference System Datum 1966. At present, because of a widespread use of GPS,
1967 (GRS 67), Geodetic Reference System 1980 (GRS 80), many local datums have been updated using the GRS 80 or
WGS 84 ellipsoid.

GRS 80 and WGS 84

Modern satellite technology has greatly improved determi-
nation of the Earth’s ellipsoid. As shown in Table 1, the semi-
major axis of the International 1924 ellipsoid is 251 m larger
than for the GRS 80 or WGS 84 ellipsoid, which represents the
current best global geodetic reference system for the Earth.
WGS 84 was designed for use as the reference system for the
GPS. The WGS 84 Coordinate System is a conventional terres-
FIG. 1. Cartoon showing the ellipsoid, geoid, and topographic trial reference system. When selecting the WGS 84 ellipsoid
surface (the landmass topography as well as the ocean
bathymetry). and associated parameters, the original WGS 84 Development
Committee decided to adhere closely to the IUGG’s approach
in establishing and adopting GRS 80.
GRS 80 has four defining parameters: the semimajor axis
(a = 6 378 137 m), the geocentric gravitational constant of the

Table 1. Examples of different reference ellipsoids and their

geometric parameters.

Semimajor axis Reciprocal of

Ellipsoid name (a in meters) flattening (1/ f )
Airy 1830 6 377 563.396 299.324 964 6
Helmert 1906 6 378 200 298.3
International 1924 6 378 388 297
Australian National 6 378 160 298.25
GRS 1967 6 378 160 298.247 167 427
FIG. 2. The elevation H above the geoid, the ellipsoid height GRS 1980 6 378 137 298.257 222 101
h, and the geoid height (undulation) N above the ellipsoid. WGS 1984 6 378 137 298.257 223 563
1662 Li and Götze

Earth including the atmosphere (G M = 3 986 005 × 108 m3 /s2 ), 1971. In the 1960s, new measurements across continents made
the dynamic form factor (J2 = 108 263 × 108 ) of the Earth ex- by precise absolute and relative gravity meters became the net-
cluding the permanent tidal deformation, and the angular ve- work of IGSN71 still in use today. A mean difference between
locity (ω = 7 292 115 × 10−11 rad/s) of the Earth (Moritz, 1980). the Potsdam datum and the IGSN71 reference has been found
Besides the same values of a and ω as GRS 80, the to be 14 mGal (Woollard, 1979).
current WGS 84 (National Imagery and Mapping Agency, Similarly, we can compare the 1967 formula to the 1980 for-
2000) uses both an improved determination of the geocen- mula in use today. The difference between the two is relatively
tric gravitational constant (G M = 3 986 004.418 ×108 m3 /s2 ) small:
and, as one of the four defining parameters, the reciprocal
(1/ f = 298.257 223 563) of flattening instead of J2 . This flat- γ1980 − γ1967 = 0.8316 + 0.0782 sin2 φ mGal.
tening is derived from the normalized second-degree zonal
gravitational coefficient (C2,0 ) through an accepted, rigorous The height correction
expression, and turned out slightly different from the GRS 80
flattening because the C2,0 value is truncated in the normal- The International Gravity Formula estimates the change
ization process. The small differences between the GRS 80 with latitude on the ellipsoid surface of theoretical gravity
ellipsoid and the current WGS 84 ellipsoid have virtually no due to an ellipsoid. The height correction accounts for the
practical consequence. change of theoretical gravity due to the station’s being located
above or below the ellipsoid at ellipsoid height h. Historically,
this height correction has been called the “free-air” correction
APPROXIMATE CALCULATION OF THEORETICAL and thought to be associated with the elevation H , not the
GRAVITY DUE TO AN ELLIPSOID ellipsoid height h. In geodesy, the “free-air” correction was
The theoretical or normal gravity, or gravity reference field, interpreted fictitiously as a reduction to the geoid of gravity
is the gravity effect due to an equipotential ellipsoid of rev- observed on the topographic surface. This has given rise to
olution. Approximate formulas are used widely even though confusion in geophysics (e.g., Nettleton, 1976, 88).
we can calculate the exact theoretical gravity analytically. Ap- As a second approximation, the height correction is given in
pendix A gives closed-form expressions as well as approximate equation (A-4). For the GRS 80 ellipsoid, we have
ones. In particular, equation (A-2) (see Appendix A) estimates δgh2 = γh − γ = −(0.308 769 1 − 0.000 439 8 sin2 φ)h
in a closed form the theoretical gravity at any position on,
above, or below the ellipsoid. + 7.2125 × 10−8 h 2 mGal. (3)
However, in exploration geophysics, a first-order formula is
The International Gravity Formula widely used, rather than this second approximation.
The conventionally used International Gravity Formula is
obtained by substituting the parameters of the relevant ref- The famous 0.3086 correction factor
erence ellipsoid into equation (A-3). Helmert’s 1901 Gravity
For the International 1924 ellipsoid, the second approxima-
Formula, and International Gravity Formulas 1930, 1967, and
tion of the height correction is (Heiskanen and Moritz, 1967,
1980, correspond respectively to the Helmert 1906, Interna-
80)
tional 1924, GRS 67, and GRS 80 ellipsoids. For example, the
1980 International Gravity Formula is (Moritz, 1980) δgh2 = −(0.308 77 − 0.000 45 sin2 φ)h + 0.000 072h 2 .
γ1980 = 978 032.7(1 + 0.005 302 4 sin2 φ Ignoring the second-order term and setting φ = 45◦ , we obtain
the first approximation of the height correction
− 0.000 005 8 sin2 2φ) mGal, (2)
δgh1 = −0.3086h mGal. (4)
where φ is the geodetic latitude.
The resulting difference between the 1980 International This is just the famous, routinely used, approximate height cor-
Gravity Formula and the 1930 International Gravity Formula rection. Again, in exploration geophysics, it is commonly called
is the (first-order) “free-air” correction and is used with the ele-
vation H rather than the ellipsoid height h.
γ1980 − γ1930 = −16.3 + 13.7 sin2 φ mGal,

where the main difference is due to a change from the Errors of approximate formulas
Potsdam gravity reference datum used in the 1930 formula to
the International Gravity Standardization Net 1971 (IGSN71) For the GRS 80 ellipsoid, as a first approximation equa-
reference. tions (2) and (4) are combined to predict the theoretical gravity
The first term of the International Gravity Formula is the at a position above (or below) the ellipsoid. The result is
value of gravity at the equator on the ellipsoid surface. Unfor- 1
γ1980 = γ1980 + δgh1 . (5)
tunately, in the 1930s, no one really knew what it was. The most
reliable estimate at that time was based on absolute gravity A second approximation is a combination of equations (2) and
measurements made by pendulums at the Geodetic Institute (3):
Potsdam in 1906. The Potsdam gravity value served as an ab- 2
solute datum for worldwide gravity networks from 1909 until γ1980 = γ1980 + δgh2 . (6)
 Correctly Understanding Gravity 1663

These two approximate formulas can be compared to the value level could be determined by regular observations at perma-
given by the closed-form formula (A-2). The two differences nent tidal gauge stations. However, one can not very accurately
are denoted as determine the elevation at a location far away from and not
1 tightly tied to an elevation datum defining mean sea level. In
1g1 = γ1980 −γ (7) practice, the geoid replaces mean sea level as a reference sur-
and face for leveling. When we level, what we really measure are
the elevations above (or below) the geoid. When geodesists or
2
1g2 = γ1980 − γ. (8) surveyors say a surface is horizontal, they really mean that it
is a surface of constant gravitational potential. So, geodesists
For an ellipsoid height of 3000 m, differences versus latitudes
have always had to measure gravity—in addition to relative
are given in Table 2. Table 3 shows differences versus ellipsoid
positions—which is why gravity historically was regarded as
heights at 45◦ latitude.
part of geodesy.
Because the differences 1g2 shown in Tables 2 and 3 are
The very early gravity work with pendulum equipment was
smaller than typical exploration survey errors, equation (A-4),
for geodetic purposes alone. Pierre Bouguer was probably the
together with the International Gravity Formula, produces a
first to make this kind of observation when he led the expedi-
sufficiently accurate approximation of the exact theoretical
tions of the French Academy of Sciences to Peru in 1735–1743.
gravity value worldwide. This equation includes the second-
Geophysical use of gravity observations started much later.
order ellipsoid height terms. For the GRS 80 ellipsoid, equa-
The first use for geological investigation may have been when
tion (A-4) becomes equation (3).
Hugo de Boeckh, who was at that time the Director of the
GEOID Geological Survey of Hungary, asked Baron Roland von
Eötvös to do a torsion balance survey over the then one-well
The geoid is a surface of constant potential energy that co- oil field of Egbell (Gbely) in Slovakia. This survey was carried
incides with mean sea level over the oceans. This definition is out in 1915–1916 and showed a clear maximum over the known
not very rigorous. First, mean sea level is not quite a surface of anticline (Eckhardt, 1940).
constant potential due to dynamic processes within the ocean. Geodesists determine the Earth’s figure (i.e., the geoid) in
Second, the actual equipotential surface under continents is two steps. First, they reduce to the geoid the gravity, observed
warped by the gravitational attraction of the overlying mass. on the actual Earth’s surface. Second, from the reduced gravity,
But geodesists define the geoid as though that mass were always they calculate the geoid undulations (i.e., the deviations from
underneath the geoid instead of above it. The main function the ellipsoid surface).
of the geoid in geodesy is to serve as a reference surface for
leveling. The elevation measured by leveling is relative to the The free-air reduction: An historical concept and requirement
geoid. of classical geodesy

GEODESY: CONVERSION OF GRAVITY TO GEOID Gravity is measured on the actual surface of the Earth. In
order to determine the geoid, the masses outside the geoid
Originally, geodesy was a science solely concerned with must be completely removed or moved inside the geoid by the
global surveying, with the objective of tying local survey nets to- various gravity corrections, and gravity must be reduced onto
gether by doing careful surveying over long distances. Geode- the geoid. Geodesists need the elevation H relative to the geoid
sists tell local surveyors where their positions are with respect when they derive the geoid from gravity.
to the rest of the world. That includes determining the elevation For a reduction of gravity to the geoid, they need the vertical
above sea level. gradient of gravity, ∂g/∂ H . Note that H ¿ a, the semimajor
axis of the ellipsoid. If gs is the observed value on the surface
Why should gravity enter into geodesy? of the Earth, then the value gg on the geoid may be obtained
as a Taylor series expansion. Neglecting all but the linear term,
Many geodetic instruments use gravity as reference. Clearly, geodesists obtain
mean sea level serves as a reference surface for leveling, and
the elevation is relative to mean sea level. In theory, mean sea gg = gs + F,

Table 2. Differences ∆g1 in equation (7) and ∆g2 in equation (8) of theoretical gravity in equation (A-2) and the two approxi-
mations in equations (5) and (6) at an ellipsoid height of 3000 m and different geodetic latitudes.

latitude 0◦ 15◦ 30◦ 45◦ 60◦ 75◦ 90◦

1g1 (mGal) −0.114 −0.192 −0.411 −0.728 −1.079 −1.363 −1.474
1g2 (mGal) 0.028 0.038 0.061 0.073 0.052 0.009 −0.013

Table 3. Differences ∆g1 in equation (7) and ∆g2 in equation (8) of theoretical gravity in equation (A-2) and the two approxi-
mations in equations (5) and (6) at geodetic latitude of 45◦ and different ellipsoid heights.

height (m) 10 100 500 1000 2000 3000 4000 5000 6000
1g1 (mGal) 0.044 0.040 0.006 −0.068 −0.326 −0.728 −1.276 −1.968 −2.805
1g2 (mGal) 0.045 0.046 0.050 0.055 0.064 0.073 0.081 0.089 0.096
1664 Li and Götze

where δV
δr = ,
∂g γ
F =− H ≈ −0.3086H mGal. (9)
∂H where γ is the theoretical gravity on the surface of the spherical
Equation (9) continues to be called the “free-air” effect. earth and δr is the departure of the geoid from a sphere.
Geodesists have assumed that there are no masses above the In general and in practice, the geoid undulations are denoted
geoid, or that such masses have been removed beforehand, so by N . They are the departure from an ellipsoid and can be
that this reduction is as though it were done in “free air”. It calculated using Stokes’ formula. Details can be found in books
is so called because, after removal of the topography by the on physical geodesy (e.g., Heiskanen and Moritz, 1967; Wahr,
complete Bouguer reduction, the gravity station is left hanging 1997).
in “free air” (Heiskanen and Moritz, 1967, 131).
In classical geodesy, geodesists employed the fiction that the
“free-air” reduction condensed the topographic masses and Geoid model
lowered the station onto the geoid, whereby the geoid became Equation (1) connects h (the ellipsoid height relative to the
a bounding surface of the terrestrial masses, and gravity gg was ellipsoid), N (the geoid undulation relative to the ellipsoid),
on the geoid (Heiskanen and Moritz, 1967, 145). Therefore, and H (the elevation relative to the geoid) (Figure 2).
Stokes’ formula can be used to calculate geoid undulations The geoid undulations range worldwide from −107 m to
from gravity. Unfortunately, geophysicists have often misun- 85 m relative to the WGS 84 ellipsoid. The primary goal of
derstood and misused this geodetic philosophy. geodesy is to develop a geoid model, which is then used to
Calculating geoid undulations from gravity connect the three values. Given N , we can compute H or h
from the other. For example, when we use GPS as a posi-
After estimating gg , gravity on the geoid, geodesists can tioning tool, we measure the ellipsoid height h. The eleva-
then derive the geoid. For a simplification, we take as example tion H can be estimated by equation (1) if we have a geoid
the spherically symmetric, rotating Earth. The derivation from model.
gravity to the geoid consists of three substeps (Wahr, 1997, 104– In general, the global or large-scale features of the geoid
108). First, calculate δg, the angular-dependent component of are expressed by a spherical harmonic expansion of the grav-
gg , by the following relation: itational potential. Its higher terms are well defined by the
GM 2 ground gravity data, and the lower terms by the satellite track-
gg ≈ 2 − ω2 a + δg. ing data. The Earth Gravitational Model 1996 (EGM96) is one
a 3
of the latest global models. It is complete through degree and
Then solve order 360. The EGM96 global geoid undulations are shown in
∂δV 2
+ δV = −δg Figure 3, and have an error range of ±0.5 to ±1.0 m worldwide
∂r a (Lemoine et al., 1998). The U.S. National Imagery and Mapping
to find δV , the angular-dependent component of the gravita- Agency recommends that it be used together with the WGS 84
tional potential. Finally, the geoid shape is given by the Bruns reference ellipsoid (National Imagery and Mapping Agency,
formula 2000).

FIG. 3. The 150 × 150 global geoid undulations produced by EGM96 (Lemoine et al., 1998). The undulations range from −107 m to
85 m. Black lines indicate coast lines.
 Correctly Understanding Gravity 1665

Short-wavelength geoid undulations In geodesy, this distinction is insignificant and can be ig-
nored in most cases because the reduction (downward contin-
The relation between spherical harmonic degree n and wave- uation) to sea level affects relatively short-wavelength anoma-
length λ of geoid undulations is: lies, which are less significant in determination of the geoid. The
2π R 40 000 000 geoid reflects very-long-wavelength density variations. This is
λ= ≈ , (10)
n n particularly true in a determination of regional or global geoid
where R = 6 371 000 m is the average radius of the Earth. undulations. But in exploration geophysics, we are interested
EGM96 extends to degree and order 360 and thus has the short- just in short-wavelength anomalies. The distinction is impor-
est spatial wavelength of 111 km. tant for us. Furthermore, it has led to an astonishing level of
At present, there exists no published truly global geoid confusion among geophysicists.
model that extends beyond degree 360 (i.e., contains a wave- In exploration geophysics, Naudy and Neumann (1965) ex-
length of shorter than 111 km). Several empirical relations have plicitly noted that the free-air and Bouguer gravity anoma-
been established to estimate how the expected power of global lies refer to the observation station. Many algorithms [e.g., the
gravity and geoid signals drops off with an increase in degree equivalent source technique of Dampney (1969)] have been
of the spherical harmonic model (Kaula, 1966; Tscherning and developed to continue gravity from an undulating observation
Rapp, 1974; Jekeli, 1978). All these relations estimate that the surface to a horizontal plane. Regardless, even in the 1980s,
global rms geoid undulation signals are less than 2 cm and some publications still referred the free-air anomaly to sea level
20 cm, when the wavelengths of undulations are 10 km and and incorrectly suggested that the measured vertical gravity
100 km, respectively. gradient should be better used to reduce observed gravity to
In a local area or nationwide, a high-resolution and accu- sea level. For example, Gumert (1985) wrote, “The free-air fac-
rate geoid model may be derived. The GEOID99 model is the tor varies significantly with horizontal position and can affect
latest one for the United States. The geoid grid with a cell the reduction of observed gravity data. Land gravity measure-
size of 1 arcminute (about 2 km) is known as a hybrid geoid ments made at varying elevation in an area of rugged topog-
model, combining many millions of gravity and elevation points raphy, processed using the standard accepted free-air factor,
with thousands of control points (i.e., GPS ellipsoid heights can produce highly erroneous maps.” Again, “Airborne grav-
on leveled bench marks). For the conterminous United States, ity gives the ability to fly multi-level lines in a survey area to
when comparing the GEOID99 model back to the same control compute the free-air factor to apply to the data.”
points, the rms difference is 4.6 cm. Its resolution may be be- The height (or improperly, “free-air”) correction should be
tween 10 and 20 km (Smith and Roman, 2001). For most of geo- made using a consistent, worldwide theoretical standard, that
physical exploration purposes, simple height conversions with is, one defined by an ellipsoid. The use of local or measured
GEOID99 in the conterminous United States can be sufficient. value is inconsistent with the objective of looking for anomalies
relative to a universal model of the earth’s gravity and is unable
CORRECTLY INTERPRETING THE FREE-AIR REDUCTION to continue observed gravity to any common level.

Heiskanen and Moritz (1967, chapter 8) defined physical

geodesy to be “classical” or “conventional” before M. S. Molo- SATELLITE ALTIMETER GRAVITY: AN EXAMPLE
densky proposed his famous theory in the 1940s, and “modern” OF CONVERTING GEOID INTO GRAVITY
thereafter. Distinguished from Stokes’ formula, Molodensky’s The primary task of geodesy is to determine the geoid from
theory says that the physical surface of the Earth can be deter- the observed gravity. However, we can go in the other direc-
mined without using the density required, for example, by the tion, as well: we can convert the observed geoid into a grav-
Bouguer correction. Heiskanen and Moritz (1967, section 8.3 ity anomaly. Satellite altimeter gravity (also called satellite-
“Molodensky’s Problem”, 293) clearly wrote: derived gravity) is such a process.
In satellite altimetry, two very precise distance measure-
The normal gravity on the telluroid [a variant of the
ments are made so that the topography of the ocean surface
geoid–authors] is computed from the normal gravity
(i.e., the geoid) is derived. First, the ellipsoid height h is mea-
at the ellipsoid by the normal free-air reduction, but
sured by tracking the satellite from a globally distributed net-
now applied upward . . . Therefore the new free-air
work of lasers and/or Doppler stations. Second, the height of
anomalies have nothing in common with a free-air
the satellite above the closest ocean surface (i.e., the eleva-
reduction of actual gravity to sea level, except the
tion H ) is measured with a microwave radar altimeter. As
name. This distinction should be carefully kept in
demonstrated in equation (1), the difference between these
mind.
two heights is just the geoid undulation N . In practice, altimeter
And on page 241, data, collected by different satellites over many years, are com-
bined to achieve a high data density and to average out sea sur-
If, as is usually done, the normal free-air gradient face disturbing factors such as waves, winds, tides, and currents.
∂γ /∂h ≈ 0.3086 mGal/m is used for the free-air re- The geoid relatively reflects deeply buried density variations.
duction, then the free-air anomalies refer, strictly In order to enhance small-scale features, the high-precision
speaking, to the Earth’s physical surface (to ground geoid is converted into gravity anomaly. The gravity anomaly
level) rather than to the geoid (to sea level) . . . How- can be computed by using inverse Stokes’ formula (the geoid-
ever, this distinction is insignificant and can be ig- to-gravity method) or by taking the derivatives of the geoid and
nored in most cases, so that we may consider 1g as using Laplace’s equation (the slope-to-gravity method; e.g., see
sea-level anomalies. Sandwell and Smith, 1997). In the real world, the conversion
1666 Li and Götze

algorithms are sophisticated, based on laws of physics, geome- tions/reductions. In particular, it is not appropriate to estimate
try, and statistics. the elevation from the ellipsoid height determined by GPS and
Anyway, there is a simple relationship between gravity then use the elevation for corrections/reductions. The extra
anomaly and geoid undulation. For two-dimensional anoma- step produces less reasonable and less significant results.
lies, an anomaly in geoid with a wavelength λ and amplitude Traditionally, geophysicists use the elevation as the vertical
N , the associated gravity anomaly 1g is given by position of the gravity station and the topographic model. The
2πγ N elevation is used in all the corrections including the height cor-
1g = , (11) rection and the complete Bouguer correction (the Bouguer
λ slab, curvature, and terrain corrections). Rigorously speaking,
where γ = 980 000 mGal, the average gravity of the Earth. This in addition we should correct observed gravity for the geoid
formula can be derived in the Fourier domain by following shape. The gravity effects due to the geoid undulations are
the work of a determination of gravity anomalies from a grid called the indirect effects (Chapman and Bordine, 1979). Li
of geoid undulations (Haxby et al., 1983). Equation (11) says and Götze (1996) explained the details of estimating the indi-
that the bump in the geoid associated with a 10-mGal gravity rect effects. For example, the indirect effect δgi h caused by the
anomaly and a wavelength of 10 km is just 16 mm. This indi- routine height correction is
cates how precise the geoid must be in order to derive gravity
anomalies useful for exploration geophysics. A number of in- δgi h = −0.3086N mGal. (12)
dependent studies (Green et al., 1998; Yale et al., 1998) show
that satellite altimeter gravity has an accuracy of about 5 mGal Thus, the indirect effect on the free-air gravity anomaly can be
and resolution of about 20 km. up to 30 mGal worldwide.
However, the amplitude of geoid undulations with a wave-
GEOPHYSICS: STATION GRAVITY ANOMALY RELATIVE length below 10 km is usually smaller than 10 cm, and the
TO THE ELLIPSOID amplitude for a wavelength of 100 km is widely smaller than
1 m. Approximately, an elevation change of 10 cm results in a
Equations (4) and (9) appear to be the same. Actually, they
change in computed Bouguer anomaly value of 0.02 mGal, and
have two important differences. First, h in equation (4) is the
1 m results in a change of 0.2 mGal. In practice, at short wave-
ellipsoid height, but H in equation (9) the elevation. Second,
lengths (say, less than 100 km), we don’t need to correct for the
equation (4) accounts for the change of theoretical gravity
geoid undulations because the geoid is very smooth, with little
due to the ellipsoid with the ellipsoid height, whereas equa-
power at those wavelengths. In petroleum exploration and in
tion (9) represents an historical endeavor of reducing gravity
particular in minerals exploration, by ignoring the geoid cor-
from the Earth’s surface to the geoid. These two differences [i.e.,
rections (i.e., the indirect effects) one is unlikely to introduce
equations (4) and (9)] distinguish geophysics from (classical)
any important relative errors across the region of investigation.
geodesy. In geophysics, we should follow equation (4) and its
implications. Use of geoid, gravity, and gravity gradient
Gravity anomaly is a station anomaly The geoid undulations, gravity anomalies, and gravity gradi-
ent changes all are due to the density variations of the Earth’s
The geophysical use of gravity is to learn about the Earth’s interior, and are transformable from one to another. The wave-
interior. We need to remove the effects of the Earth’s irregular lengths that the gravity gradient, gravity, and geoid dominate
(nonellipsoidal) surface. In principle, this means that we should or concentrate range gradually from short (tens of meters) to
compare the observed gravity to that of ellipsoidally-produced long (thousands of kilometers). The geoid undulations are used
theoretical gravity values at each observation station. Their to study the global or very regional problems such as the man-
difference is just the gravity anomaly. The free-air anomaly is tle convection. On the contrary, the gravity gradient is better
the difference between the observed gravity, without terrain- used to investigate short wavelength effects for engineering,
related corrections, and the theoretical gravity. The complete environmental, or mining problems.
Bouguer anomaly is the difference between the observed grav-
ity with the complete Bouguer correction (the Bouguer slab, SUMMARY
curvature, and terrain corrections) and the theoretical gravity.
Geodesy uses gravity to determine the geoid. Geodesists
Both the free-air and Bouguer gravity anomalies are located at
must reduce the observed gravity from the actual surface of
the gravity station. We must conduct a continuation process in
the Earth to the geoid (mean sea level). In the gravity cor-
order to obtain the gravity responses on the geoid or another
rections/reductions, geodesists use the elevation instead of the
surface/level. As an example, in a continuation to sea level of
ellipsoid height.
the ground Bouguer gravity anomaly in the Central Andes, the
Geophysics uses gravity to study the Earth’s interior. The
correction value can reach 30% of the maximum magnitude of
gravity anomaly is the difference between the observed gravity
the station anomalies (Li and Götze, 1996).
and the theoretical gravity predicted from the ellipsoid. The
The ellipsoid height, or the elevation plus the geoid gravity anomaly is located at the observation station after the
height correction and other routine corrections/reductions are
In geophysics, the gravity anomaly is the difference be- applied. In principle, the ellipsoid height should be used in
tween the observed gravity and the theoretical gravity pro- positioning and in all data corrections/reductions. In practice
duced by the ellipsoid. The geophysical gravity anomaly can and in minerals and petroleum exploration, use of the elevation
be calculated simply by using the ellipsoid height h instead rather than the ellipsoid height hardly introduces significant
of the elevation H in positioning and in all necessary correc- errors because the geoid is very smooth. However, it should
 Correctly Understanding Gravity 1667

not be recommended as a routine procedure to derive the ele- Haxby, W. F., Karner, G. D., LaBrecque, J. L., and Weissel, J. K.,
vation from the ellipsoid height determined by GPS and then 1983, Digital images of combined oceanic and continental data sets
and their use in tectonic studies: EOS, 64, 995–1004.
use the elevation for corrections/reductions. Heiskanen, W. A., and Moritz, H., 1967, Physical geodesy: W. H.
The theoretical gravity on, above, and below the ellipsoid Freeman and Co.
Jekeli, C., 1978, An investigation of two models for the degree variances
surface can be calculated by a closed-form formula. Its approx- of global covariance functions: Ohio State University, Department
imation by the International Gravity Formula and the height of Geodetic Science and Surveying, Report 275.
correction including the second-order terms is typically accu- Kaula, W. M., 1966, Theory of satellite geodesy: Blaisdell Publishing
Co.
rate enough worldwide. Lakshmanan, J., 1991, The generalized gravity anomaly: Endoscopic
microgravity: Geophysics, 56, 712–723.
Lemoine, F. G., et al., 1998, The development of the joint NASA GSFC
ACKNOWLEDGMENTS and National Imagery and Mapping Agency (NIMA) geopotential
model EGM96: Technical Paper NASA/TP-1998-206861.
A part of this work was finished when X.L. worked as a re- Li, X., and Götze, H.-J, 1996, Effects of topography and geoid on grav-
search fellow at Freie Universität Berlin, financially supported ity anomalies in mountainous areas: The Central Andes as an ex-
ample: Institut für Geologie, Geophysik und Geoinformatik, Freie
by the Alexander von Humboldt Stiftung. Writing of this tuto- Universität Berlin.
rial has been largely stimulated by the discussions at the gravity Moritz, H., 1980, Geodetic Reference System 1980: Bulletin
and magnetic user group grvmag-l during May and June 2000. Géodésique, 54, 395–405.
National Imagery and Mapping Agency, 2000, Department of Defense
Comments from and discussions with Richard O. Hansen and World Geodetic System 1984: Its definition and relationship with
Maurice D. Craig were helpful in improving the tutorial. We local geodetic systems: Technical Report NIMA TR8350.2, Third
also thank associate editor David A. Chapin and reviewers Edition.
Naudy, H., and Neumann, R., 1965, Sur la définition de l’anomalie de
Alan T. Herring and Dhananjay N. Ravat for their valuable Bouguer et sec conséquences pratiques: Geophy. Prosp., 13, 1–11.
suggestions and comments. Nettleton, L. L., 1976, Gravity and magnetics in oil prospecting:
McGraw-Hill Book Co.
Sandwell, D. T., and Smith, W. H. F., 1997, Marine gravity anomaly from
REFERENCES Geosat and ERS-1 satellites: J. Geophys. Res., 102, 10039–10054.
Smith, D. A., and Roman, D. R., 2001, GEOID99 and G99SSS:
Chapman, M. E., and Bordine, J. H., 1979, Considerations of the indi- One arc-minute models for the United States: J. Geodesy, in
rect effect in marine gravity modeling: J. Geophys. Res., 84, 3889– press.
3892. Tscherning, C. C., and Rapp, R. H., 1974, Closed covariance expres-
Dampney, C. N. G., 1969, The equivalent source technique: Geophysics, sion for gravity anomalies, geoid undulations, and deflections of the
34, 39–53. vertical implied by anomaly degree variance models: Ohio State
Eckhardt, E. A., 1940, A brief history of the gravity method of prospect- University, Department of Geodetic Science and Surveying, Report
ing for oil: Geophysics, 5, 231–242. 208.
Green, C. M., Fairhead, J. D., and Maus, S., 1998, Satellite-derived gra- Wahr, J., 1997, Geodesy and gravity: Class notes: Samizdat Press.
vity: Where we are and what’s next: The Leading Edge, 17, 77–79. Woollard, G. P., 1979, The new gravity system—Changes in interna-
Gumert, W. R., 1985, Advantages of continuous profiling airborne tional gravity base values and anomaly values: Geophysics, 44, 1352–
gravity surveys: Proceedings of the International Meeting on Po- 1366.
tential Fields in Rugged Topography, Institut de Géophysique, Uni- Yale, M. M., Sandwell, D. T., and Herring, A. T., 1998, What are the
versité de Lausanne, 16–18. limitations of satellite altimetry?: The Leading Edge, 17, 73–76.

APPENDIX A
THEORETICAL GRAVITY DUE TO AN ELLIPSOID

The theoretical gravity is the gravity effect due to an equipo- and a and b are the semimajor and semiminor axes of the ellip-
tential ellipsoid of revolution. Approximate formulas are used soid, respectively; γe and γ p are the theoretical gravity at the
widely. In fact, we can calculate the theoretical gravity at any equator and poles, respectively; and φ is the geodetic latitude.
position on, above, or below the ellipsoid surface using closed-
form expressions. Closed-form formula: Gravity above and below the surface of
the ellipsoid
Closed-form expression: Gravity on the surface of the ellipsoid
The theoretical gravity at any ellipsoid height h and
The theoretical gravity on the surface of the ellipsoid is given any geodetic latitude φ (Figure A-1) can also be given by a
by the formula of Somigliana (Heiskanen and Moritz, 1967, 76): closed-form formula. Starting from the general formula of
1 + k sin2 φ Heiskanen and Moritz (1967, 67–71), Lakshmanan (1991) de-
γ = γe q , (A-1) rived the formula and published a result containing typo-
1 − e2 sin2 φ graphic errors. Li and Götze (1996) repeated the derivation
and corrected the errors, obtaining
where
1 ω2 a 2 Eq 0
½ µ
1 1
¶
GM
k=
bγ p
− 1; γ = 02 2
+ 02 2
sin2 β 0 −
aγe W b +E (b + E )q0 2 6
r ¾
a 2 − b2 − ω2 b0 cos2 β 0 , (A-2)
e= is the first eccentricity;
a2
1668 Li and Götze

where and R = r 002 /E 2 , D = d 002 /E 2 , r 002 = r 02 + z 02 , d 002 = r 02 − z 02 , r 0 =

p a cos β + h cos φ, z 0 = b sin β + h sin φ, and tan β = b/a tan φ.
E = a 2 − b2 is linear eccentricity,
s Approximate formula for the latitude correction
b02 + E 2 sin2 β 0
W = , The conventional latitude correction is a second-order series
b02 + E 2
expansion of equation (A-1) (Heiskanen and Moritz, 1967, 77):
¶
1
µ
b02 b0
µ ¶µ ¶
0 −1 E ∗ 2 2
γ = γe 1 + f sin φ − f 4 sin 2φ , (A-3)
q = 3 1+ 2 1 − tan − 1,
E E b0 4
with
3b2
¸
3b
·µ ¶
1 −1 E γ p − γe
q0 = 1 + 2 tan − , f∗ = (gravity flattening),
2 E b E γe
p 1 5
b0 = r 002 − E 2 cos2 β 0 , f 4 = − f 2 + f m,
2 2
s a−b
f = (flattening of the ellipsoid),
r
0 1 R 1 R2 D a
cos β = + − + − ,
2 2 4 4 2 and
ω2 a 2 b
m= .
GM

Approximate formula for the height correction

The height correction accounts for the change of theoret-
ical gravity due to the station being located above or below
the ellipsoid at ellipsoid height h. As a second approximation
(Heiskanen and Moritz, 1967, 79), a Taylor series expansion
for the theoretical gravity above the ellipsoid with a positive
direction downward along the geodetic normal to the reference
ellipsoid is
· ¸
2 3
γh = γ 1 − (1 + f + m − 2 f sin2 φ) h + 2 h 2 .
a a
The difference γh − γ (i.e., the height correction) is
FIG. A-1. A station above an ellipsoid surface. The ellipsoid ¶ ¸
3γe
· µ
has the semimajor axis a and semiminor axis b. The position of 2γe 5
γh − γ = − 1 + f + m + m − 3 f sin2 φ h + 2 h 2 .
station P relative to the ellipsoid is defined by ellipsoid height a 2 a
h and geodetic latitude φ. The angle β is called the reduced
latitude. (A-4)
See ERRATA for this Figure
 Errata 997

GEOPHYSICS, VOL. 67, NO. 3 (MAY-JUNE 2002); DOI 10.1190/1.1489656 031203GPY

To: “Ellipsoid, geoid, gravity, geodesy, and geophysics,” X. Li and H. -J. Götze (Geophysics, 66, 1660-1668).

The authors thank Dr. Nico Sneeuw of the University of ellipsoid. All the formulae given in Appendix A are correct.
Calgary for pointing out a graphical error. The reduced lati- There is also a typographic error in the dynamic form
tude β was incorrectly defined in Figure A-1, which should factor J2, in the first paragraph on page 1662.
be replaced by the figure shown below. However, this
J2 = 108 263 x 10-8 not 108 263 x 108.
graphical error was not introduced into the derivation of the
closed-form expression for the theoretical gravity due to an

Figure A-1. A station above a reference ellipsoid surface. The reference ellipsoid has a semi-major axis a and a semi-minor
axis b. The position of station P relative to the reference ellipsoid is defined by ellipsoid height h and geodetic latitude φ. The
ellipsoid through point P has the same linear eccentricity as the reference ellipsoid. The reduced latitude β is a geocentric lat-
itude of point Q, which is the vertically projected point, on a sphere of radius a, of station P’s normal projection on the refer-
ence ellipsoid surface.