You are on page 1of 5

B35

A Fast Least-squares Method for Inverse


Modeling of Gravity Anomaly Profiles due Simple
Geometric-shaped Structures
K.S. Essa* (Cairo University)

SUMMARY
An inversion technique using a fast least-squares method is developed to estimate, successively, the shape
factor (q-parameter), the depth (z-parameter) and the amplitude coefficient (A-parameter) of a buried
structure using normalized residual anomalies obtained from gravity data. By defining the anomaly value
at the origin and the anomaly value at different points on the profile (N-value), the problem of shape factor
estimation is transformed into a problem of finding a solution of a non-linear equation of the form f(q)=0.
Knowing the shape factor and applying the least-squares method, the depth is estimated by solving a
nonlinear equation of the form ψ(z) = 0. Finally, knowing the shape factor and the depth, the amplitude
coefficient is determined in a least-squares way using a simple linear equation. This technique is
applicable for a class of geometrically simple anomalous bodies, including the semi-infinite vertical
cylinder, the infinitely long horizontal cylinder, and the sphere. The technique is tested and verified on a
theoretical model with and without random errors. It is also successfully applied to real data from mineral
exploration in India, and the interpreted shape and depth parameters are in good agreement with the known
actual values.

Near Surface Geoscience 2012 – 18th European Meeting of Environmental and Engineering Geophysics
Paris, France, 3-5 September 2012
Introduction

The important targets in the interpretation of gravity data are to estimate the nature of the source
(shape factor), the depth and the amplitude coefficient which is related to the density and the
parameters of the structure. It is known that the gravity data interpretation is non-unique where
different subsurface causative targets may yield the same gravity anomaly; however, a priori
information about the geometry of the causative target may lead to a unique solution (Roy et al.
2000). Several numerical methods have been developed to estimate the nature of the sources and the
depth of the sources such as: using characteristic points and distances (Essa, 2011), transformation
techniques (Mohan et al., 1986; Shaw and Agarwal, 1990), a simple formula approach (Essa, 2007a),
graphical method (Roy et al., 1999), least-squares minimization approach (Abdelrahman and
Sharafeldin, 1995; Abdelrahman et al., 2001), Euler deconvolution technique (Zhang et al., 2000), use
of moving average residuals (Abdelrahman et al., 2006), and use of horizontal gradient residuals
(Essa, 2007b).
A new fast three least-squares algorithm for gravity inversion is developed which estimate the shape
factor parameter (q-parameter) using a non-linear least-squares sense. Once, q is known, the depth (z-
parameter) estimation problem is transformed into the problem of finding also a solution of a non-
linear function f(z)=0. The solution is obtained by minimizing a function in the least-squares way.
Finally, after knowing the q-parameter and z-parameter, the amplitude coefficient (A-parameter)
parameter is estimated in a least-squares sense using a linear formula. Using the entire measured data
makes the results produced more reliable and realistic, and helps minimize the uncertainties due to the
non-uniqueness and ill-posedness of the inverse problem solution. The proposed method has been
applied to synthetic example with and without random noise, and to a real dataset from mineral
exploration in India.

The Method

The gravity anomaly produced by a simple geologic structure along a profile over the body is given
by
zm
g ( x i , z , q) = A , i = 1, 2 , 3 ,....., k (1)
(x + z 2 )q
2
i
where z is the depth of the source, q is a factor related to roughly to the shape of the sources, xi is the
position coordinate. At the origin (xi = 0), the equation (1) gives the following relationship:
A
g (0) = , (2)
2q − m
z
Using equation (2), we obtain the following normalized gravity anomaly form:
q
⎛ 2 ⎞
⎜ z ⎟
F( x i , z , q ) = ⎜ ⎟⎟ , (3)
⎜ x + z2
2
⎝ i ⎠
g (x i , z.q)
where F( x i , z , q ) = . Again, for all shapes, equation (3) gives the following value at xi = ±N
g (0 )
q
⎛ z2 ⎞
T=⎜ ⎟ , N = ±1, 2 , 3 ,...... ( 4)
⎜ N2 + z2 ⎟
⎝ ⎠
From equation (4), we obtain the following equation in the depth (z):
1
2 N2T q
z = . ( 5)
1
1− T q
Substituting equation (5) into equation (3), we obtain the following equation in the shape factor (q):
q
⎛ N 2 P(q) ⎞
⎜ ⎟
F( x i , q ) = ⎜ , (6 )
⎜ x + P(q)(N − x ) ⎟⎟
2 2 2
⎝ i i ⎠
1
where P(q) = T q . The unknown shape factor (q) in equation (6) can be obtained by minimizing:

Near Surface Geoscience 2012 – 18th European Meeting of Environmental and Engineering Geophysics
Paris, France, 3-5 September 2012
M 2
ϕ(q) = ∑ [L(x i ) − W(x i , q)] , (7 )
i=1
q
⎛ N 2 P(q) ⎞
where L(xi ) denotes the normalized observed gravity anomaly at xi and W (x i , q) = ⎜⎜ 2 2 2 ⎟
⎟ .
⎝ x i + P(q)(N − x i ) ⎠
Setting the derivative of φ(q) to zero with respect to q leads to
M
f(q) = ∑ [L(x i ) − W(x i , q)]W * (x i , q) = 0 , (8)
i=1
d
where W * ( x i , q) = W ( x i , q) . Equation (8) can be solved for q using the standard methods for solving
dq
nonlinear equations (e.g. Mustoe and Barry, 1998), and its iteration form can be expressed as:
q f = f(q j ) , (9 )

where qj is the initial shape factor and qf is the revised shape factor; qf will be used as the qj for the
next iteration. The iteration stops when qf − q j ≤ e , where e is a small predetermined real number close
to zero. Substituting the computed shape factor (qc) as a fixed parameter in equation (3), we obtain
qc
⎛ 2 ⎞
⎜ z ⎟
F( x i , z ) = ⎜ ⎟⎟ . (10)
⎜ x2 + z2
⎝ i ⎠
Again, applying the least-squares method, the unknown depth (z) in equation (10) can be obtained
from the following non-linear equation:
1
⎡ M x i2L(xi ) ⎤ 2q c
⎢ ∑ ⎥
⎢ i = 1 (x 2 + z 2 )qc + 1 ⎥
zc = ⎢ i ⎥ . (11)
⎢M x i2 ⎥
⎢ ∑ 2 2 2q c + 1 ⎥
⎣⎢ i = 1 (x i + z ) ⎦⎥

where L(xi) is the normalized gravity anomaly data. Substituting the computed shape factor (qc) and
depth (zc) in equation (1) as fixed parameters, we obtain:
z cm
V(x i ) = A , (12)
(x i2 + z c2 ) qc

where V(xi) is the observed gravity data. Finally, applying the least-squares method to equation (12),
the unknown amplitude coefficient Ac can be determined from
M V(x i )

i = 1 ( x 2 + z c2 ) qc
i
Ac = . (13)
M z cm

i = 1 ( x 2 + z c2 ) 2 qc
i
We then measure the goodness of fit between the observed and computed gravity data for each set of
solutions. The standard error (µ) is used in this paper as statistical preference criteria in order to
compare the observed and calculated values. This µ is given by the following mathematical
relationship:
k
[
∑ g(xi ) − g c (xi ) 2 ]
μ= i=1 , (14)
k
where g(xi) is the observed gravity values and gc(xi) is the calculated gravity values. The minimum µ
used as a criterion for determining the optimal model parameters (q, z, A) of the buried structure.

Synthetic example

Numerical example has been studied using the above-mention technique. Figure 1 shows the residual
gravity data due to a model defined by the parameters A = 300 mGal*km, z = 3 km and q = 1.0. The
related normalized gravity anomaly is shown in Figure 2. Equations (8), (11) and (13) were applied to
the normalized residual anomaly profile (Fig. 2). Table 1 (it will be presented at conference) shows in
detail the obtained inversion results. One can see that the proposed algorithm recovered the true

Near Surface Geoscience 2012 – 18th European Meeting of Environmental and Engineering Geophysics
Paris, France, 3-5 September 2012
model parameters for all used N-value. This data set was then corrupted by an additive 10% random
error, and inverted. The percentage of error is 2.8%, 7.8% and 6.7% for q-parameter, z-parameter and
A-parameter, respectively. The percentage of error in any model parameter is still less than the 10%
random noise added to the gravity data. This illustrates the desirability of the present approach
compared to the conventional methods which use only few points from the supplied data while
computing the model parameters of the buried structure. The detailed inverse results of the free-noise
and noisy data will be presented at the conference.
1.0
100 Model Parameters: Model Parameters:
q = 1.0 q = 1.0
z = 3.0 km z = 3.0 km
A = 300.0 mGal*km 0.9
90 A = 300.0 mGal*km

Normalized residual gravity anomaly (mGal)


0.8
80
Residual gravity anomaly (mGal)

0.7
70

0.6
60

0.5
50

0.4
40

0.3
30

0.2
20

10 0.1

0 0.0

-10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10
Horizontal distance (km) Horizontal distance (km)

Figure 1 The residual gravity anomaly of a Figure 2 The normalized gravity anomaly of
horizontal cylinder model (q = 1.0, z = 3 km, a horizontal cylinder model.
and A = 300 mGal*km).

Field Example

The residual gravity anomaly over a manganese deposit near Nagpur, India (Reddi et al.,
1995) was shown (Fig. 3). This profile has a length of 333 m, the gravity profile was digitized at
interval of 55.55 m. The proposed inverse technique has been applied to the observed data to estimate
the q-parameter, z-parameter and A-parameter using the normalized field of the observed gravity data
(Fig. 4). Then we computed the standard error (µ) between the observed values and the values
computed from estimated parameters q, z and A for each N-value. The results are shown in Table 2 (it
will be presented at conference) for the cases of N-value. Also we computed the set of mean values
and the optimum set (µ = 0.005 mGal) is given at N = 111.1 m. The best-fit-model parameters are q =
1.15, z = 59.18 m and A = 7.91 mGal*m (Fig. 3). This suggests that the shape of the buried structure
resembles a 2-D horizontal cylinder model buried at a depth of 59.2 m. The shape and the depth to the
center of the ore body obtained by the present method agree very well with those obtained from Roy
(2001).

Conclusion

A fast least-squares approach has been developed to estimate the appropriate nature of the source (q-
parameter), the depth (z-parameter) and the amplitude coefficient (A-parameter) of a buried structure
from the normalized gravity anomaly data a long profile. The inverse fast algorithm has been derived
for fast gravity quantitative interpretation for geometrically simple anomalous bodies, such as a 3D
semi-infinite vertical cylinder, a 2D infinitely long horizontal cylinder, and a 3D sphere. The
suggested method is automatic and it can use all the observed gravity data in estimating of these three
parameters; previous techniques have typically used only a few points, distances, standardized curves,
and nomograms. The suggested algorithm is found to be stable and can estimate the gravity
parameters with a reasonable accuracy even when the observed data is contaminated with noise, and
the origin of the gravity structure is approximately determined. The method has been successfully
tested on a synthetic example with and without random errors, and successfully applied to a field

Near Surface Geoscience 2012 – 18th European Meeting of Environmental and Engineering Geophysics
Paris, France, 3-5 September 2012
example from India. The estimated gravity inverse parameters are found in a good agreement with the
known published values.
0.35 1.0

Observed anomaly
Calculated anomaly 0.9
0.30
0.8

Normalized gravity anomaly (mGal)


0.25 0.7
gravity anomaly (mGal)

0.6
0.20
0.5

0.15 0.4

0.3
0.10

0.2

0.05
0.1

0.0
0.00
-160 -120 -80 -40 0 40 80 120 160
-160 -120 -80 -40 0 40 80 120 160 Horizontal distance (m)
Horizontal distance (m)
Figure 3 The residual gravity anomaly Figure 4 Normalized gravity anomaly data
over a manganese deposit near Nagpur, over a manganese deposit near Nagpur,
India. India.

References

Abdelrahman, E.M. and Sharafeldin, S.M. [1995] A least-squares minimization approach to depth
determination from numerical horizontal gravity gradients. Geophysics, 60, 1259–1260.
Abdelrahman, E.M., Abo-Ezz, E.R., Essa, K.S., El-Araby, T.M. and Soliman, K.S. [2006] A least-
squares variance anaylsis method for shape and depth estimation from gravity data. Journal of
Geophysics and Engineering, 3, 143-153.
Abdelrahman, E.M., El-Araby, T.M., El-Araby, H.M. and Abo-Ezz, E.R. [2001] A new method for
shape and depth determinations from gravity data. Geophysics, 66, 1774-178.
Essa, K.S. [2007a] A simple formula for shape and depth determination from residual gravity
anomalies. Acta Geophysica, 55, 182-190.
Essa, K.S. [2007b] Gravity data interpretation using s-curves method. Journal of Geophysics and
Engineering, 4, 204-213.
Essa, K.S. [2011] A new algorithm for gravity or self-potential data interpretation. Journal of
Geophysics and Engineering, 8, 434-446.
Mohan, N.L., Anandababu, L. and Roa, S. [1986] Gravity interpretation using the Melin transform.
Geophysics, 51, 114–122.
Mustoe L.R. and Barry, M.D.J. [1998] Mathematics in Engineering and Science. Wiley, New York.
Reddi, A.G.B., Murthy, B.S.R. and Kesavanani, M. [1995] A Compendium of Four Decades of
Geophysical Activity in Geological Survey of India. GSI Special Publication No. 36,Geological
Survey of India.
Roy L. [2001] Short note: Source geometry identification by simultaneous use of structural index and
shape factor. Geophysical Prospecting, 49, 159-164.
Roy L., Agarwal, B.N.P. and Shaw, R.K. [1999] Estimation of shape factor and depth from gravity
anomalies due to some simple sources. Geophysical Prospecting, 47, 41–58
Roy, L., Agarwal, B.N.P. and Shaw, R.K.[2000] A new concept in Euler deconvolution of isolated
gravity anomalies. Geophysical Prospecting, 16, 559–75.
Shaw, R.K. and Agarwal, P. [1990] The application of Walsh transforms to interpret gravity
anomalies due to some simple geometrical shaped causative sources: A feasibility study. Geophysics,
55, 843-850.
Zhang, C., Mushayandebvu, M.F., Reid, A.B., Fairhead, J.D. and Odegard, M.E. [2000] Euler
deconvolution of gravity tensor gradient data. Geophysics, 65, 512–520.

Near Surface Geoscience 2012 – 18th European Meeting of Environmental and Engineering Geophysics
Paris, France, 3-5 September 2012

You might also like