You are on page 1of 6

GEOPHYSICS, VOL. 63, NO. 4 (JULY-AUGUST 1998); P. 1125–1130, 7 FIGS., 2 TABLES.

Editor’s Note:
Most discussion papers in GEOPHYSICS are very brief and relegated to the section on “Discussions”. However,
in the present case, we are publishing two short notes by renowned experts on the topic of Bouguer corrections for
gravity data. It is an enlightened discussion where two prominent researchers have a slight disagreement. The first
paper “Errors in the Total Bouguer Reduction” by Manik Talwani explains that differences between the Bullard
corrections and those published by Tom LaFehr result from the use of different gravitational constants. However,
LaFehr claims that the Bullard corrections are still slightly in error. These short notes are placed in sequence so
that the readers of GEOPHYSICS may form their own opinions.
—Larry Lines
Editor

Short Note

Errors in the total Bouguer reduction

Manik Talwani∗

INTRODUCTION A = 0.1118h (h being the height in meters), which corresponds


to G = 6.6635 × 10−8 cgs units, and he used this value to arrive
This short note discusses various errors in the Bouguer re-
at values for B in such a way that the sum A + B gave the cor-
duction that have been discussed recently in the literature. It
rect value in Cassinis’ tables. In current usage, G = 6.67 × 10−8 ,
also discusses two other errors arising out of the “indirect ef-
which yields A = 0.111897h, and correspondingly a different
fect” and the uncertainty in the value of the universal constant
value for B. This difference is what led Hensel and LaFehr to
of gravitation, G, that previously have not been discussed ex-
question the accuracy of the Cassinis and Bullard/Swick tables.
tensively in the literature.
Table 1 demonstrates the agreement of the spherical cap for-
GRAVITY EFFECT OF A SPHERICAL CAP AND mula given by Bullard and Swick with the exact determination
THE BULLARD B CORRECTION (Talwani, 1973; Cassinis et al., 1937).
Thus, in fact, neither the Cassinis tables (which were con-
LaFehr (1991b, 1992) and Hensel (1992) discuss the gravity structed in response to a request by the International Asso-
effect of a spherical cap. They infer that the values given by ciation of Geodesy and which are considered a landmark in
Bullard (1936) and Swick (1942) for the Bullard B correction gravity calculations) nor the Bullard and Swick values for the
(which is the difference between the gravity effect of a spherical Bullard B correction are in error. LaFehr’s (1991b, 1992) and
cap to 166.735 km and the effect of an equally thick infinite Hansel’s (1992) remarks stem from a misunderstanding of the
flat plate) are in error (see Figure 1). Hensel (1992) wonders way in which Bullard obtained the Bullard B correction.
whether Bullard used a value for the radius of the Earth that
is greatly in error (5975 km instead of 6371 km) or whether STANDARDIZATION OF BOUGUER REDUCTION
Cassinis’s tables (Cassinis et al., 1937), which Bullard used, are
in error. It has been argued that for gravity surveys for the purpose
Actually, there is no error in values given by Bullard (1936) of exploration, reduction parameters for obtaining Bouguer
and Swick (1942), and there is a simple explanation why anomalies should be standardized by choosing an outer radius
Bullard’s values for B are different from values currently in to Hayford zone 0 (166.735 km), a density of 2.67 gm/cm3 ,
use. and should include terrain corrections (LaFehr, 1991a). Stan-
Bullard (1936) used the correct value of the spherical cap dardization has obvious merits, but we need to consider the
from Cassinis’ tables and expressed it by the formula: following questions:
Bouguer spherical cap = A + B, (1)
1) What is the effect of ignoring topography beyond
where A is the flat-plate Bouguer factor and B is the Bullard 166.735 km?
B correction. The quantity A involves the universal constant 2) Is correction to 166.735 km really necessary for very local
of gravitation G. Although the Cassinis tables adopt a value surveys (where all the stations are within a few kilome-
of 6.67 × 10−8 for G, Bullard used the then prevalent value ters)?

Manuscript received by the Editor February 05, 1998; revised manuscript received February 05, 1998.

Houston Advanced Research Center, The Woodlands, TX 77381 and Rice University MS 126, Dept. of Geology and Geophysics, 6100 Main St.,
Houston, TX 77005-1892. E-mail: Manik@owlnet.rice.edu.
°c 1998 Society of Exploration Geophysicists. All rights reserved.

1125
1126 Talwani

3) Are circular zones the best method for calculating the of the spherical cap, we consider station 1 surrounded by a
Bouguer effect (including terrain correction)? spherical cap of radius r 1. The effect of ignoring topography
beyond r 1 is greatest for the topography immediately beyond
Figure 2 can be used to visualize the kind of error caused by r 1. We have taken a block 20 × 20 × 5 km lying at a distance
ignoring the topography beyond 166.735 km. We illustrate the between r 1 and r 1 + 20 km from the station and calculated its
error with an example. Imagine a station situated in the center effect at the station. Then, we moved the station a distance r 2
of a circular plateau of height 2000 m and radius 1000 km, with in the opposite direction, to station 2, recalculated the effect
the elevation being zero at all points beyond 1000 km. The of the block and took a ratio of the effects at stations 1 and 2.
error caused by considering a spherical cap that extends only If the ratio is close to unity, the implication is that the effect
to 166.735 km is approximately 17 mGal. of bodies immediately outside the spherical cap does not vary
This error obviously depends on the elevation of the station much between stations 1 and 2, and thus can be separated from
and the surrounding topography, but it is clear that the error can local effects. We have carried out the calculation for three val-
be considerable. The only way to eliminate the error completely ues of r 1 (30, 150, and 500 km), and various values of r 2. Thus,
is to consider the topography over the entire Earth, which is as one sees in Figure 3, that if the Bouguer spherical cap has a
obviously impractical in the context of the present methods of radius of 30 km, moving the station 16 km away from a block
determining the Bouguer anomaly for surveys taken for ex- lying between 30 and 50 km reduces its effect to 0.45 of the
ploration purposes. The error in ignoring topography beyond previous value. If, however, the radius were 150 or 500 km, the
166.735 km may not be troublesome, however, for such sur- effect is reduced to 0.9 or 0.99, respectively. As expected, for a
veys if the error varies slowly over the extent of the survey. given distance r 2 the rate of reduction of the effect is greatest
Intuitively, this is so if the stations in the survey are relatively for r 1 = 30, and least for r 1 = 500. If we arbitrarily choose a
close together. A slowly varying error can be considered a part cutoff ratio of 0.9 in Figure 3, we conclude that all the stations
of a regional anomaly and, hence, eliminated during interpre- within the survey should lie within 2.5 km for a cap of 30 km,
tation. On the other hand, if a survey includes stations that are and should lie within about 20 km for a cap of 150 km, but could
far apart (that is, the interstation distances are of the order of extend to several hundred kilometers for a cap of 500 km.
magnitude of the spherical cap radius), this error will be differ- The assumptions we have made in the above calculations
ent at different stations and will lead to errors in interpretation. are quite arbitrary. They only suggest in a general way that for
To obtain a rough indication of the variation of the gravita- exploration surveys extending over long distances, ignoring to-
tional effect of topography of bodies lying beyond the radius pography beyond 166.735 km may lead to problems in inter-
pretation. We do not suggest that the radius of the Bouguer
Table 1. Comparison of spherical cap values.∗ cap be increased to large values. That would be impractical for
a number of reasons. Later in the paper, we suggest a different
(1) (2) (3) (4) scheme for calculating Bouguer anomalies.
Height of Spherical cap Spherical cap Spherical cap
spherical cap exact formula Cassinis table Bullard/Swick
(m) (mGal) (mGal) (mGal)
0 0 0 0
100 11.333 11.294 11.38
200 22.658 22.642 22.66
300 33.976 33.962 33.94
500 56.592 56.551 56.6
1000 113.008 112.99 113.0
1500 169.246 169.20 169.2
2000 225.309 225.28 225.3
2500 281.195 281.18 281.2
3000 336.904 336.90 336.9
3500 392.437 392.38 392.4
4000 447.794 447.79 447.8

Column (2) gives the exact value of the gravity effect of a
spherical cap (using the exact solution in a form given by
Talwani, 1973) at a station at its top (see Figure 1). Column
(3) gives the corresponding values from the Cassinis table
(Cassinis et al., 1937; Lejay, 1947). Column (4) gives the corre-
sponding values using the Bullard’s (1936) formula and using
the values of Bullard B given by Swick (1942). The values agree
in all cases within about 0.04 mGal.

FIG. 1. The spherical cap to 166.735 km and the infinite


Bouguer flat plate. The difference in the gravity effect of the
two at the station constitutes the Bullard B correction. The FIG. 2. The gravity effect of spherical caps of various thick-
curvature of the Earth is greatly exaggerated in this figure. nesses and various radii minus the effect of an infinite flat plate.
Errors in the Total Bouguer Reduction 1127

For very local surveys, a few kilometers in extent, a small reasoning is flawed. Since the topography to 166.735 km is very
radius for the zone to which terrain corrections are applied has nearly the same for the two stations, we should not be calculat-
the benefit that a flat Earth approximation (Heiskanen and ing the effect of two spherical caps, but rather the effect of the
Vening Meinesz, 1958) can be used. It is customary to use the same topography at the two stations. This is implicitly achieved
infinite flat-plate value for the Bouguer correction and then by making a terrain correction to the two spherical caps. When
apply a flat-plate terrain correction to a zone with a radius of this is done, it is easy to see that for very closely spaced stations
about 30 km. As discussed earlier, if the effect of topography at different elevations, the quantity to be concerned with in the
outside the 30-km zone varies slowly for the stations in the local above example is not the gradient of the Bullard B correction,
survey, this effect can be accounted for by suitably choosing the but is the difference in the effect of a 1-km spherical cap at
regional anomaly. its surface (113.00755 mGal) and of the same 1-km spherical
However, one other factor has to be considered. LaFehr cap at a point 1 m above it (113.00684 mGal). This difference
(1991b) stated that for closely spaced stations at different ele- equals −0.71 µGal, which can be expressed as a gradient of
vations, the variation (with elevation) in the Bullard B cor- −0.71 µGal/m (Figure 5). Thus, the gradient to be considered
rection needs to be taken into account. His point can be is −0.71 µGal/m, rather than +0.75 mGal/m. However, it is in-
illustrated as follows: Consider a plateau at about 1 km ele- appropriate to take even this gradient into consideration. The
vation, but with two stations with an elevation difference of elevation above sea level, in reality, does not equal zero beyond
1 m (say, one station is at an elevation of 1 km, and the other at 166.735 km, and the value of gradient cited above is an arti-
1.001 km). The infinite flat-plate Bouguer values are 111.89662 fact arising out of the truncation of the cap at 166.735 km. By
and 112.00852 mGal respectively, the difference being increasing the radius of the spherical cap to, say, 5000 km, this
111.90 µGal (Figure 4). But if the spherical cap to 166.735 km gradient can be greatly reduced; but to make terrain correc-
was considered, the difference would be 113.12020 − tions for a local survey to a distance of 5000 km obviously is
113.00755 mGal, i.e., 112.65 µGal. The difference between the absurd.
two differences is +0.75 µGals. In other words, the vertical A solution to the vertical gradient problem for very local
gradient of the Bullard B correction is +0.75 µGals/m at an surveys lies in simply using the flat plate approximation; that
elevation of 1 km. Thus, according to LaFehr, even for very is, using an infinite flat Bouguer plate and making terrain cor-
local surveys, the Bullard B correction needs to be applied for rections to, say, 30 km. The vertical gradient for an infinite flat
a spherical cap extending to 166.735 km. However, this line of plate is zero. Any errors that this leads to simply arise out of the
neglect of distant topography (in this case, beyond 30 km) and
would be assumed to vary slowly through the survey. Barring

FIG. 4. Gravity effect of a 1-m-thick plate lying on top of spher-


ical caps of various thicknesses and various radii. Scale on the
left gives the value in µGal; scale on the right is obtained by sub-
tracting the effect of a 1-m-thick infinite flat plate. The effect is
numerically equal to a gradient expressed in microgals/meter.
FIG. 3. Ratio of gravity effect at stations 1 and 2. Station 1 As discussed in the text, the use of this gradient is flawed be-
lies at a distance r 1 from a spherical trapezoid 20 × 20 × 5 km. cause it implies that the topography to the boundary of the cap
Station 2 lies at a distance r 1 + r 2 from the trapezoid. is different for the two nearby stations.
1128 Talwani

very special situations, the gradient would be very small if dis- a distant square (or, more appropriately, a spherical trapezoid)
tant topography were being considered, and thus the infinite can be represented by a line element. The gravity effect of a ra-
flat plate serves as a good approximation. dial line element on a spherical earth was given by Kukkamäki
Do we specially need to be concerned with the vertical gra- (1955) and Talwani (1973). Obviously, very distant small
dient for surveys that are not local? By carrying out terrain squares can be combined into bigger squares. Table 2 gives the
corrections up to the radius of the spherical cap, the vertical size of a square for this calculation as a function of the distance
gradient is automatically accounted for. However, as discussed to and beyond the square for which any error caused by com-
above, the truncation of the cap gives rise to a value of the bining smaller squares into bigger squares is less than 1%. Thus,
gradient that is incorrect. Thus, paradoxically, the error con- beyond 30◦ , the 50 × 50 squares can be combined into 10◦ × 10◦
cerning the vertical gradient does not arise out of its neglect. squares; between 2◦ and 3◦ distance, the 50 × 50 square should
Rather, an incorrect value arises out of the truncation of the be combined to 0.5◦ × 0.5◦ squares, and so on. At distances
cap. In general, the smaller the radius of the cap, the larger will under 0.3◦ (roughly equal to 37 km), 50 × 50 squares are not
be the error. accurate enough, and the terrain obtained locally (generally in
conjunction with the gravity survey itself) should be used.
A NEW SYSTEM FOR BOUGUER CORRECTIONS Besides being able to consider the entire Earth if necessary,
this method has two other advantages. First, since the contri-
Surveys for exploration often extend to several tens of kilo-
bution from very distant topography varies slowly with the ge-
meters (and that may lead to errors if a cap radius of 166.735 km
ographical position as well as elevation of a station, tables can
is chosen); but instead of extending the Bouguer correction by
be constructed that will list the value of the contribution as a
the spherical cap method to say, 500 km, we suggest that a differ-
function of the latitude, longitude, and elevation of the station.
ent system of making Bouguer corrections may be more useful.
The value of the contribution of distant topography at any sur-
The method is similar to the method suggested by Heiskanen
vey point will not have to be calculated for every station, but
(1953) for making topographic and isostatic corrections in out-
can be obtained easily from the tables by suitable interpolation.
side numbered Hayford zones. The new system would make
Second, by choosing fixed squares rather than compartments
use of the fact that fairly reliable elevation (and depth) values
defined by radial and circumferencial lines, the contribution
exist over 50 squares (obtainable from the National Geophysi-
from a particular square can be upgraded easily without redo-
cal Data Center, Boulder, Colorado) all over the Earth. For the
ing the entire calculation if newer and better elevation values
purpose of considering the Bouguer effect on a spherical Earth,
are available for that square.
INDIRECT EFFECT

The “indirect” effect arises from the fact that while station
elevations are referred to the geoid, the International Gravity
Formula is referred to a reference ellipsoid, and the two may
not coincide. Chapman and Bodine (1979) discuss the “indi-
rect” effect for sea measurements, but the effect is also appli-
cable to land measurements.
In Figure 6, the elevation at the gravity station is the dis-
tance AB. N , the geoid height above the reference ellipsoid,
is given by N = BC. If the geoidal height N remained con-
stant, the difference in reference surfaces would not have any
consequence for exploration-related gravity reductions. If, on
the other hand, the geoid height changes rapidly, as it does in
mountainous areas, the discrepancy would have to be taken
into account. It is easily seen that the indirect effect involves
a “free-air” and a “Bouguer” reduction if the geoid is above

Table 2. Size of squares used for calculations.∗


Size of Square Approximate
representable by number of
Distance Line Element squares
Greater than 30◦ 10◦ × 10◦ 64 750
20◦ –30◦ 5◦ × 5◦ 60
◦ ◦ ◦ ◦
7 –20 2.5 × 2.5 250
3◦ –7◦ 1◦ × 1◦ 150
2◦ –3◦ 0.5◦ × 0.5◦ 75
1◦ –2◦ 0.25◦ × 0.25◦ 160
◦ ◦ 0 0
0.3 –1 5 ×5 500

FIG. 5. The difference in the gravity effect of spherical caps of The size of a square (column 2) that can be represented by a
various thickness and various radii directly over the center of line element at a distance in column 1, for gravity calculations
the cap and 1 m above that point. Numerically, this difference error less than 1%. Column 3 gives the number of squares at
equals a gradient expressed in microgals/meter. distances in column 1.
Errors in the Total Bouguer Reduction 1129

the ellipsoid. The “free-air” and “Bouguer” corrections have equal to 0.4 mGal. For steeper gravity gradients and longer
to be made in the same sense as these corrections are made wavelength anomalies, the indirect effect will be correspond-
to account for the elevation of the station above the geoid. ingly larger. Regional anomalies are generally assumed to be
If the geoid is below the ellipsoid, the correction would be in linear rather than sinusoidal. It can be shown that if H varies
the opposite sense. Making the flat plate approximation, the by a over a distance λ, the corresponding change in geoid
correction is (0.3086 − 0.1119) = 0.1967 mGal/m (0.3096 is the height is approximately aλ/2g. If the horizontal gravity value
free-air factor and 0.1119 is the flat-plate Bouguer factor for a H changes by 50 mGal over a distance of 200 km (i.e., a gra-
density of 2.67). dient of 0.25 mGal/km exists over 200 km), the corresponding
The variation of geoid height can exceed 5 m over a distance change in geoid elevation is 5 m and the indirect effect will
of 100 km, but is usually less than that. However, in moun- amount to about 1 mGal. Geoid undulations of this magnitude
tainous regions, gradients of 2 m/100 km are not uncommon. are actually present in central Europe.
The value of the indirect effect corresponding to anomalies of
simple geometry can be calculated as follows. THE VALUE OF G, THE UNIVERSAL
The observed value of gravity is the vertical component of GRAVITATIONAL CONSTANT
gravity, V . Since the same mass distribution is obviously re- Arguably, G is the most difficult physical constant to measure
sponsible for both components V and H (the horizontal com- because gravity is the weakest of the four known fundamen-
ponent of gravity), let us assume for this calculation that both tal forces, and it is impossible to shield delicate measurements
components have similar magnitudes. Let the value of gravity from the gravitational influences of buildings and other nearby
(for which we will calculate the geoid undulation) be sinusoidal objects. Experiments by Luther and Towler (1982) have led to
and be described by an amplitude of a mGal and a wavelength the accepted value of 6.672610 ± 0009 × 10−8 cgs units. How-
of λ km. If we assume that both components of gravity are ever, Schewe (1995) cites three newer determinations reported
equal, then the variation of the horizontal component of grav- in 1994 from the University of Wuppertal, Germany; Measure-
ity with distance x is given by ments Standard Laboratory, New Zealand; and the German
x Bureau of Standards, RTB Braunschweig. The three determi-
H = a sin 2π . (2) nations, respectively, are 6.668510 ± 0007, 6.665610 ± 0006, and
λ
6.715410 ± 0006 × 10−8 . Thus, it appears that the uncertainty in
The deflection of vertical d (for small angles) is given by the value of G, and consequently in the Bouguer correction,
H a x lies between 1% and 0.1%. For a spherical cap at an elevation
d= = sin 2π . (3) of 2 km, the error in the Bouguer correction lies between 2 and
g g λ 0.2 mGal.
The amplitude N of the geoid is then obtained by integration
DISCUSSION OF ERRORS
of d with respect to x:
Most of the errors discussed above are small. Even more
aλ aλ
N= , and peak-to-trough amplitude 2N = , importantly, they are slowly changing, which implies that they
2πg πg can be lumped with regional anomalies and separated from
(4) residual anomalies. One aspect of reductions that needs to be
considered seriously is the outer limit to which Bouguer re-
or
ductions and terrain corrections are carried out. The errors

2N = , (5) resulting from a truncation of a finite distance depend on the
3000 topography involved and also on the interstation distance, i.e.,
where a is in milligals and λ is in kilometers. the separation distance between stations in the same survey
If we consider a sinusoidal anomaly with amplitude a = compared to the distance to which the corrections are carried
25 mGal and λ = 250 km (Figure 7), the peak-to-trough am- out. A long-term solution would be to adopt a different system
plitude of the sinusoidal geoid fluctuation will be 2N ≈ 2 m, for reductions instead of circular zones—a system that would
with an indirect effect difference at the peak and the trough depend on squares of various sizes whose boundaries are de-
termined by latitudes and longitudes.

FIG. 7. A sinusoidal horizontal gravity variation gives rise to a


FIG. 6. The station elevations are referred to the geoid (or the sinusoidal geoid undulation with the same wavelength. If the
sea level surface), but the International Gravity Formula is re- gravity variation has an amplitude of 25 mGal and a wavelength
ferred to the ellipsoid. The difference gives rise to the “indirect” of 250 km, the amplitude of the sinusoidal geoid undulation is
effect in gravity reduction. approximately 2 m (see text).
1130 Talwani

Although it is often stated that the assumptions and er- 2) The above criteria are suggested for mineral-related grav-
rors involved in the reduction can be accounted for during ity surveys and are not applicable to highly precise, very
interpretation, it is not necessarily easy to do so in practice. local gravity surveys for archaeological and other pur-
For example, consider a survey carried out over a space of, say, poses. For these surveys, very accurate measurements and
10 × 10 km. The Bouguer reduction carried to 166.735 km as- accurate reductions (but carried out to only a relatively
sumes a constant density down to sea level of 2.67 gm/cm3 . In small distance, say, 30 km) are necessary.
theory, any deviation from the constant density can be included
in the interpretation; in practice, however, the interpretation REFERENCES
will be confined to the area of the survey and not extended to
Bullard, E. C., 1936, Gravity measurements in East Africa: Phil. Trans.
166.735 km. Roy. Soc. London, 235, 486–497.
The inaccuracies mentioned above in gravity reductions are, Cassinis, G., Dore, P., and Ballarin, S., 1937, Tavole fondamentali per
fortunately, often small in comparison with the uncertainties la riduzione dei valori osservati della gravita: Publicazione dell’ In-
stituto di Geodesia 13.
inherent in the interpretation of gravity anomalies. Errors in Chapman, M. E., and Bodine, J. H., 1979, Considerations of the indirect
knowledge of topography and uncertainties in the separation effect in marine gravity modeling: J. Geophys. Res., 84, 3889–3892.
of regional and residual anomalies are important. But perhaps Heiskanen, W. A., 1953, Isostatic reductions of the gravity anomalies
by the aid of high speed computing machines, Publ. Isos. Inst. IAG
most important, and generally outweighing by far any errors (Helsinki), no. 28.
in the reduction of gravity values, are the uncertainties in den- Heiskanen, W. A., and Vening Meinesz, F. A., 1958, The Earth and its
sities of rock bodies giving rise to the gravity anomalies. In the gravity field: McGraw-Hill Book Co.
Hensel, E. G., 1992, Discussion on: “An exact solution for the gravity
interpretation of gravity anomalies, density contrasts come into curvature (Bullard B) correction” by T. R. LaFehr: Geophysics, 57,
play. They seldom exceed 0.5 gm/cm3 , and it is seldom possible 1093–1094.
Kukkamäki, T. J., 1955, Gravimetric reductions with electronic com-
to establish them to an accuracy better than 10%. This fact sug- puters, Publ. Isos. Inst. IAG (Helsinki), no. 30.
gests that if the gravity reductions are carried out to an order of LaFehr, T. R., 1991a, Standardization in gravity reduction: Geophysics,
magnitude better accuracy, that is to 1%, we may have a satis- 56, 1170–1178.
——— 1991b, An exact solution for the gravity curvature (Bullard B)
factory situation. With this in mind, I would like to suggest the correction: Geophysics, 56, 1179–1184.
following in developing guidelines for the accuracy of gravity ——— 1992, Discussion on: “An exact solution for the gravity cur-
reductions: vature (Bullard B) correction” by T. R. LaFehr: Geophysics, 57,
1094.
Lejay, R. P., 1947, Developpements moderns de la gravimetrie:
1) The accuracy goal of reductions should be 1% or Gauthier-Villars.
0.1 mGal, whichever is greater. The 1% figure makes the Luther, G. G., and Towler, W. R., 1982, Redetermination of the New-
tonian gravitational constant G, Phys. Rev. Letters, 48, 121–123.
accuracies in reduction an order of magnitude higher than Schewe, P. R., 1995, Physics update, Science News, June, 263.
accuracies in interpretation. The 0.1-mGal figure is rele- Swick, C. H., 1942, Pendulum gravity measurements and isostatic re-
vant to the great effort involved in making gravity mea- ductions: U.S. Coast and Geodetic Sur. Spec. Publ. 232.
Talwani, M., 1973, Computer usage in the computation of gravity
surements as well as reductions with higher accuracy—an anomalies, in Bolt, B. A., Ed., Geophysics: Methods in computa-
effort that may not be justifiable in extra costs. tional physics 13: Academic Press Inc., 343–389.

You might also like