BOUNDARY VALUE PROBLEMS

Overview ofboundary value problems

Boundary-value problems give the potential outside the earth, where there are no
masses and where the potential, satisfying Laplace's equation, is harmonic. The
determination of potential fromn the masses is a direct problem while the
determination of the masses from potential is an inverse problem. Such inverse
problems have no direct solutions and are encountered in geophysical prospecting by
gravity measurements: underground masses are inferred from the disturbances of the
gravity field.
We define Dirichlet's problem, or the frst boandary-vale problem of potential theory, as
follows: Given an arbitrary function on a surface S, to determine a function V which
is harmonic either inside or outside S and which assumes on S the values of the
prescribed function. This can be solved not only for the sphere but also for any
sufficiently smooth boundary surface.

In Neumann's problem, or the second boundary-value problem of potential theory, the normal
derivative ôV/ôn is given on the surface S, instead of the function V itself. The
normal derivative is the derivative along the outward-directed surface normal n to S.
derivative
In thethird bonndary-value problem, a linear combination of V and its normal
hV + k
On
(2.1)

is given on S.
The third boundary-value problem is particularly relevant to the physical geodesy,is
because the determination of the undulations of the geoid from gravity anomalies
just such a problem.
Geodetic boundary-value problem
The geodetic boundary-value problem comprises the determination of the surface of
the Earth and its external gravity field from observations on or close to the Earth's
Surface. The surface to be determined is either the geoid (Stokes problez) or the
physical surface of the Earth and the quasi-geoid (Molodensky problen)
21
In space we have the well-known fact that gravity vector g and the gravity
potential
(geopotential) Ware related by

g=grad W=
dy' ôz.
(2.2)
which shows that the force g is the gradicnt vector of the potential.
Let S be on the Earth's topographic surface and let W and g be the geopotential and
the gravity vector on this surface. Then there exists a relation
g= f(S, w)
(2.3)
the gravity vector g on S is a function of the surface S and the geopotential W on it.
This can be seen on the following way. Let the surface S and the geopotential W on
S be given. The gravitational potential V is obtained by subtracting the potential of
the centrifugal force , which is simple and perfectly known:
V=W- ¢
(2.4)

The potential V outside the Earth is a solution of Laplace's equation AV =0 and

consequently harmonic. Thus, knowing V on S, we can obtain V outside S by solving
the first boundary-value problem, the Dirichlet's problem, which is practically always
uniquely solvable at least if V is sufficiently smooth on S. After having found V as a
well
function in space outside S, we obtain the gravitational force grad V. Adding the
known and simple vector of centrifugal force, we obtain the gravity vector g outside,
by continuity, on S.
can be
This is precisely what (2.3) means. The modern general concept of afunction
uniquely
explained as a rule of computation, indicating the given S and W on S, we can
but rather a
calculate g on S. Note that f is not a function in the elementary sense
may
"non-linear operator", but we disregard this for the moment. Therefore we
formulate:

22
 (1) Molodensky's boundary-value problem is the task to determine S, the
Earth's surface,
if g and W on it are given. Formally, we have to solve (2.3) for S:

S= F(g. W)
(2.5)
that is, we get geometry from gravity,.
(2) GPS boundary-value problem. Since we have GPS at our disposal, we can
consider S as known, or at least determinable by GPS. In this case, the
geometry S is known, and we can solve (2.3) for W:
W= Fz(S,g)
(2.6)
that is, we get potential fromgravity. This leads to a method to replace leveling, a tedious
and time-consuming old-fashioned method, by GPS leveling, a fast and modern
technique.
In spite of all similarities, we should bear in mind a fundamental difference: (2.6)
solves afixed-bonndary problem (boundary Sgiven), whereas (2.5) solves afre-boundary
probler. the boundary S is a priori unknown ( free"). Fixed-boundary problems are
usually simpler than frec ones.
boundary-value
NB: The geodetic boundary-value problem resembles the third
given a linear
problem of potential theory, namely to determine a harmonic function
surface. It
combination of the function and its normal derivative on a bounding
be known:
differs from the classical problem, as the bounding surface is supposed to
not represent
free boundary-value problem. In addition, observed gravity data do rather refer
of the Earth, but
potential derivatives with respect to the physical surface
of the position vector cannot be
to the plumb lines. Finally, the horizontal component consequently the
determined with sufficient accuracy from gravimetric data,
the determination of heights.
geometric part of the problem is generally restricted to
geometry of the
Withrapid progress in satellite positioning and satellite altimetry, the
still increasing accuracy; the
Earth's surface now can be assumed to be known with
then is the external gravity
only remaining unknown at the boundary value problemn formulated, which employs
potential. Hence, a fixed boundary-value problem can be
the second (Nemann)
gravity disturbances as boundary values. This corresponds to
23
boundary-value problem of potential theory, which is to determine a harmonic
function from its derivative given on the boundary surface. Finally, a mixed altimetric
gravimetric boundary-value problem may be set up, taking into account that - in
addition to the harmonic coefficients derived from satellite orbit analysis or satellite
gradiometry altimetric geoid heights and gravity anomalies are the main data sets
available on a global scale.

24
 ACQUISITION OF GRAVITY DATA
Gravity measurements
Gravity data can be acquired through measurement of absolute or relative gravity.
Absolute gravity measurements

Absolute gravity measurement involves detemination of g from the fundamental

acceleration quantites of length and time. Methods used include the free-fall, rise-and
fall, and pendulum methods.
Through the integration of ~ g for a free-fall experiment, one obtains the relation
Z= Zo + |ot +
(3.1)
between the path z and the fall time t. The constants of integration Zo and |o
represent the position and velocity of the body at t = 0. If the body falls through at
least three planes (Fig. 3.1a), Zo and |, can be climinated. We have

g=2
(z3 - Z)(t, - t)- (z, - Z)(t3 -ti)
(t3 - t)(ty - t)(t3 - tz)
(3.2)

a) z (m), APEX b)
0,4 -

S1y
0,2
0,2
AZ
FAL
0,4 Vo 3mla

0,2 t(o) 0,2 0,4 0,6 tlo)

Figure 3.1: Distance timediagram: (a) free-fall method, (b) symmetrical rise-and-fall
method.

25
 For a symmetric
rise-and-fall (Fig. 3.1 b), it is sufficient to measure the crossing times
t, tz, tg, t4 at any two planes (seperated by the
distance Az). (3.1) then yields
8Az
g=
(ta - t)2 - (ty t,)?
(3.3)
If the relative error of a free-fall experiment is not to exceed 10
to t1 x 10- ms), a falling distance of 0.5 m and a
(corresponding
corresponding falling tíme of
0.3 s yields auragy requirments of t Snm and t0.2 ns. This can be achieved by
interferometric measurements with laser light and electronic time measurement. All
instruments in use today employ the simultaneous length and time measuremnt, with
a Michelson interferometer. Two corner cube reflectors are the fundamental parts of the
interferometer (Fig. 3.2).

JLONG PERIOD
SUSPENSION
FALLING
CORNER CUBE
REFLECTOR
REFERENCE
CORNER
CUBE
REFLECTOR

LASER
BE AM
MIRROR SPUTTER
PHOTO
DETECTOR

MIRROR

the free-fall method

Figure 3.2: Principle of Michelson interferometer used in
reference, the other can be moved in the
One of the reflectors is fixed and serves as a and
and a reference beam,
vertical. By spliting the laser light into a measurement
after a parallel reflection, light interferences occu. The zero
superimposing them
(wave-length A= 632.8 nm). Ifn
crossing of the fringe signal have a distance of A/2
(zero crossings) are counted, this corresponds to a falling distance of.
pulses
26
 Time is measured with an atomic frequency standard
(rubidium normal) after a large
preset number of fringes has occurred (e.g., after 10000 fringes,
which leads to
approximately 60 time measurements during one drop over 0.5 m). Consequently,
(3.1) has to be evaluated by least squares adjustment (on-line
processing).
The pendulum method is based on the measurement of the
period and length l of a freely
swinging pendulum. For a mathematical pendulum (point mass suspended on a
weightless wire), we have the differential equation
lp + gsin Í = 0,
(3.4)
where p is the phase (Fig. 3.3a). Integration leads to the period of oscilation

T = 21t
16

(3.5)
The amplitude Po generally remains less than 30'.

02
mg
b

Figure 3.3;,Absolute pendulum principle; (2) mathematical pendulum (b) reversible

pendulum.

27
The mathematical pendulum can notbe strictly realized.
However, (3.4) and (3.5) hold
as well for the physical pendulum, if lis replaced by the reduced length of
the pendulum.
J
l, =
ma

(3.6)
Here, / is the moment of inertia with respect to the axis of rotation 0, anda is the
distance between this axis and the center of mass S. For a reversible pendulum (Fig.
3.3 b), L, may be determined as the distance between the points of suspension 0, and
02 which give the same period of oscilation T.
Relative gravity measurement
The measurement of a difference in gravity Ag by the direct or indirect observation of
one of the two accleleration quantities of time or length keeping the other one fixed,
is known as relative" gravity measurement; it can be performed with considerably
more ease than the absolute measurement of g. A distinction is made between
pendulum and spring gravimeter neasurements.

In a pendulum measurement, the periods of oscilation T, T, of the same pendulum

are measured at two points P, (91) and P2(gz). Assuming invariance of the length l
of the pendulum, it follows from (3.5), after some simple manipulation that
T, -TË (T; -T)²
Ag12 = 92 - 91 =-2g1
T, T
(3.7)
"Thus the length measurement is not required here, as opposed to the case of absolute
determination, so that an arbitrary physical pendulum (generally, l = 0.25 m) can be
used. With the invention of the spring gravimeter which is more exact and
economical, the pendulum measurements have lost their importance (the time
required to makea measurement including set up and disassembly of the pendulum
apparatus is approximately one day, for the gravimeter is about five minutes).
The (relative) spring gravimeter is based on the principle of a spring balance. The
equilibrium position of a mass m is observed as it is influenced by the acceleration of
gravity and by the counterforce of the elastic spring, If gravity
changes the spring
28
 length will also change in order to reach static equilibrium again. According to
law, the strain is proportional to the stress for small elongations. We Hooke's
between translational systemns (rarely uscd) and the rotational systems.
differentiate
In a ranslational system (vertical spring balances), the condition of
equilibrium (Fig, 3.4
a) is given by

mg - f(l- l,) = 0
(3.8)
where f is thespring constant and l(lo) is the length of the spring with load (without
load). Differentiating (3.8) furnishes a lincar relationship between the change in
gravity Ag and the observed difference in length Al:

Ag = m

(3.9)
a) c)

b)

1f0-4)
T(y" a)measuring
spring
m m
ng

mg mg

Figure 3.4: Spring balance gravimeter principle; (a) vertical spring balance, (b) lever
torsion spring balance, and (c)general lever spring balance
In order to assess the gravity changes with a relative accuracy of t10-, length
changes of a 0.1 mn long spring have to be determined to t1 nm.

Rotational ystems (lever spring) consists of a lever which supports a mass m and rotates
about an axis 0. Equilibrium is produced through a horizontal torsion spring or

29
 through a vertically or oblique acting restoring spring. For the lever torsion spring
balance (Fig.3.4b), the equilibrium of the torques yields
mga cosa - T(y + a) = 0
(3.10)
where a=length of the lever, a = angle between the horizntal and the lever, T=
torsion constant, Y= pretension angle of the spring. For a = 0, a linear relationship
exists between Ag and the angle of deflection Aa

Ag = ma
-Aa =Aa
(3.11a)
At the general lever spring balane, the spring counterforce f(l L) can act under an
arbitrary angle on the lever carrying the mass m (Fig. 3.4c). The line connecting the
rotation axis O with the point where the spring is mounted deviates by an angle 8
from the vertical. The equilibrium condition for the torques reads

mga sin(a + 8) + fbd l-lo sin a = 0

(3.11b)
The sensivity of this non-linear system may be significantly improved by
approximating the torques of gravity and of elastic spring (astatization). With a zero
length (o = 0) spring, this is achieved with asmall angle 8,and a s 90°. In order to
obtain the +10-8 relative accuracy, we now need (at a = 0.1 m, a + 8= 90°,
8 = 100') the displacement to be measured with t2 um. Compared to the linear
system the sensitivity is increased by the factor 2000.
Gravity measurenent on the Ocean and in the Air
The application of gravimnetric method and development of Earth models
presuppOses the existance of the complete knowledge of the gravity field. Therefore
gravity measurements must also be made on the oceans. Results are obtained more
rapidly with airborne gravimeter measurements.

Ordinary gravimeters that are used on land, but wich have been built into a pressure
and water protected case, and equipped with remote control and recording devices are
30
 known as underater graimeters. They are transported and
lowered to the ocean floor for assembled on a ship and
measurements.
If gravity measurements are made on a ship or in an
airplane, then one must solve the
problem leveling the instrument and separating the perturbing
of
actual gravity field.
acelerations from the
These disturbing accelerations vary with peak periods of 5 to 10 son a ship
and 1to
300 s on an airplane. The corresponding amplitudes can attain and exceed 0.5 ms
on the oceans and 0.05 ms for airborne measurements. While the horiznatal
component of the disturbing potential can be reduced by instrumental or
methodological measures, the vertical conponents enters completely into the
observations.

The gravity meter may be gimbal-suspended or placed on a gyro-stabalized platform. In the

first case, the resultant of g and the horizontal disturbing acceleration is measured
(Browne effect), thereby requiring a correction that is often difficult to obtain. Today,
the instruments are exclusively mounted on a stabalized platform (the uncertainty in
leveling the instrument is +1). The oflevehng effect caused by the remaining incination
can
generally be neglected.
The measuring systems must be strongly dampened due to the large verical
accelerations. The short-periodic perturbations are filtered out by taking averages over
sufficiently long intervals of time (1 to 5 minutes). This low-pass fltering smoothes
the data, and yields mean gravity values over the distances travelled during the
averaging time interval.
Reductions

Intoroduction

Gravity g measured on the physiscal surface of the Earth must be distinguished from
the normal gravity y reffering to the surface of the ellipsoid. To refer g to sea level, a
reduction is necessary. Since there are masses above sea level, the reduction methods
differ depending on the way in which these topograpbical masses are dealt with. Gravity
reduction is cssentially the same for gravity anomalies Ag and gravity disturbance Sg.
Gravity reduction serves as a tool for three main purposes:

31
 determination of the geoid,
interpolation and extrapolation of gravity,
" investigation of the earth's crust
Only the first two are of adirect geodetic nature. The third is of interest to theoitical
geophysicists and geologists, who study the gencral structure of the crust, and to
exploration geophysicists.
The use of Stoke's formula for the determination of the geoid requires that gravity
anomalies Ag represent boundary values at the geoid. This implies two conditions:
first, gravity g must refer to the geoid; second, there must be no masses outside the
goid. Hence, figuratively speaking, gravity reductions consists of the following steps:
1. the topographical masses outside the geoid are completely removed or shifted
below sea level;
2. then the gravity station is lowered from the Earth's surface (point P) to the
geoid (point Po, see Fig. 3.5)

earth's surface
H

geoid
Po

Figure 3.5: Gravity reduction

The first step requires knowledge of the density of the topographic masses, which is
somewhat problematic.
By such a reduction procedure certain irregularities in gravity due to differences in
height of the stations are removed so that interpolation and even extrapolation to
unobserved areas become casier.
Free-air reduction

32
 For a theoretically correct reduction of gravity to the geoid, we need ôg/ôH, the
vertical gravity gradient. Ifg is the observed value at the surface of the Earth, then
the value go at the geoid may be obtained as a Taylor expansion:
ôg
OH-H
(3.12)
where H is the height between P, the gravity station Aabove the geoid, and Po, the
corresponding point on the geoid (l"ig. 3.5). Suppose there are no massea bove the
geoid and neglecting all term but the lincar one, we have
go = g +F
(3.13)
where

F= H
OH
(3.14)
Is the free-air reduction to the geoid. Note that the assumption of no masses above the
geoid may be interpreted in the sense that such masses have been mathematically
removed beforehand, so that this reduction is indeed carried out in free air".
For many practical purposes it is sufficient to use instead of dg/OH the gradirent of
normal gravity (associated with the cllipsoidal hcight h) y/ôh, obtaining,
ßy
F=-H= +0.3086 H
ah
(3.15)
for H in meters.

Bouguer reduction
The objective of the Bouguer reduction of gravity is the complete removal of the
topographical masses, that is, the masses outside the geoid.
The bonguer plate
33
 Assume the area around the gravity station P to be
3.6), and let the masses between the completely flat and horizontal (Fig.
geoid and the Earth's surface have a constant
density . The attarction due to the so called Bouguer plate is
given by
Ap = 2GQH
(3.16)
With the standard density = 2.67 g cm this becomes
A = 0.1119 H [mgal]
(3.17)
for Hin metes.

Po
Figure 3.6: Bouguer plate
Removing the plate is equivalent to subtracting its attraction from the observed
gravity.
To continue and complete our gravity reduction,we must now apply the free-air
reduction F as given in (3.15). This combined process of removing the topographical
masses and applying the free-air reduction is called complete Bouguer reduction. Its result is
Bouguer gravity at the geoid:
IB =9- Ag + F
(3.18)
With the assumed numerical values, we have Bouguer gravity at Po
gB = g -0.1119 H+0.3086 H =g+0.1967 H
(3.19)
Since ga now refers to the geoid, we obtain genuine gravity anomalies by subtracting
the normal gravity Yreferred to the ellipsoid:
34
 Agg = gB Y
(3.20)
They are called Bonguer anomalies.
Terrain correction

This simple procedure is refined by taking into account the deviation of the actual
topography from the Bouguer plate of P (Fig. 3.7). This is called terrain correction or
topographical corection.
A earth's surface

B P
H-HP
.- Bouguer plate

Hp

geoid
Po
Figure 3.7: Terrain correction

At A the mass surplus Am+, which attracts upwards is removed, causing g at P to

increase again.
increase. At B the mass deficiency Am_ is made up, causing g at P to
The terrain correction is alvays positive.
use of a template
For practical determination of terrain correction, Ac, is through
comnponents, AA, are then
made of various compartments, the effects of individual
summed up as

A, =)AA
(3.21)
graity
By adding the terrain correction A, to (3.18), the refined bonguer
ge =9 - Ag t A, + F
(3.22)

35
 is obtained. The Bouguer reduction and the
are called refined or simple, depending corresponding bouguer anomalies Age
on whether the terrain correction has been
applied or not.
In practice it is convinient to separate the Bouguer reduction into
effect of a Bouguer
plate and the terrain correction, because the amount of the latter is
usually much less.
Even for mountains 3000 m in height, the terrain correction is only of the
order of
50mgal.
Unifed approach
It is also possible tocompute the total effectof topographical masses,
A, = Ag - At, (3.23)
in one step by using colunms with base at sea level (Fig. 3.8), again subdividing the
terrain by means of a template.

Hp

Po

Figure 3.8: Bouguer reduction

NB: Note the difference between A, the attraction of the topographical masses, and
the terrain correction A,.

Then

AA

36