What to expect from Euler deconvolution estimates for isolated sources

Felipe Ferreira de Melo (Observatório Nacional)* and Valeria Cristina Ferreira Barbosa (Observatório Nacional)

Copyright 2017, SBGf - Sociedade Brasileira de Geofísica

Fourier methods. However, Fourier methods are also
th
used to calculate horizontal gradients (Blakely, 1996) and
This paper was prepared for presentation during the 15 International Congress of the
Brazilian Geophysical Society held in Rio de Janeiro, Brazil, 31 July to 3 August, 2017. these methods’ stability depend on the signal-to-noise
level. Pašteka et al. (2009) regularized the derivatives in
Contents of this paper were reviewed by the Technical Committee of the 15th
International Congress of the Brazilian Geophysical Society and do not necessarily Fourier domain to keep them stable in cases with low
represent any position of the SBGf, its officers or members. Electronic reproduction or signal-to-noise ratio. Other methodology to calculate
storage of any part of this paper for commercial purposes without the written consent
of the Brazilian Geophysical Society is prohibited. gradients in Euler deconvolution is the equivalent layer
___________________________________________________________________ principle (Leao and Silva, 1989) done by Barbosa et al.
Abstract (1999), Silva and Barbosa (2003) and Melo et al. (2013).

Euler deconvolution is a technique based anomaly Thompson (1982) noticed the relationship between the
measurements, its derivatives and structural index. The correct structural index and correct depth estimates. He
objective is to estimate the base level of the data, observed that the correct structural index produces the
horizontal and vertical positions of the source, usually smallest spread of solutions. Thus, he proposed a
assuming a tentative structural index. Derivatives are criterion to identify the correct structural index based in
calculated using different techniques and different the clustering of depth solutions. Reid et al. (1990)
methodologies exist to indicate the correct structural followed this line of tight depth estimates and proposed
index. Most of the approaches define the correct depth their criterion. On the other hand, Barbosa et al. (1999)
estimates based on structural index estimated. We noticed that the correct structural index is the one that
modeled anomalies magnetized by vertical and inclined yield the minimum correlation, in modulus, between the
directions, generated by an isolated magnetic monopole estimated base level and the observed total-field anomaly
and calculated its analytical derivatives in order to show for 2D data, Melo et al. (2013) used this tecnique in 3D
what expect from Euler deconvolution estimates for single data. Silva and Barbosa (2003) gave the theoretical basis
sources. We then analyzed estimates over the source for selecting solutions assuming a null base level in their
influence area and at the borders. These examples formulation. They defined the behavior of horizontal
clearly show an outstanding pattern for depth and base estimates at the borders of the anomaly. Over the source,
level estimates when the correct structural index is used. they showed that horizontal estimates form plateaus of
The estimates at these cases, and only at these cases, constant values and these estimates are correct
are constant over the anomaly position. For these cases, regardless of the structural index. While vertical estimates
this pattern can clearly identify the correct structural form plateaus of values close to the correct one but they
index, related to the nature of the source. are only correct when the correct structural index is used.

Introduction In this paper we generated synthetic anomalies provided

by a single isolated monopole (Telford et al., 1990) and
Euler deconvolution was proposed in the nineties (Reid et do not calculate the derivatives. Instead, we used
al., 1990) and became widely accepted as a semi- analytical derivatives of a monopole in Euler
automatic interpretation method for magnetic and deconvolution in order to avoid any error or undesired
gravimetric data interpretation. The technique is based in effect that generated by numerical derivatives. Assuming
Euler homogeneity equation and applies for single point three tentative structural indices, we then analyze the
sources related to geological structures (Reid and estimates using wrong and correct structural indices.
Thurston, 2014). It relates anomaly measurements, its These estimates are landmarks in the analysis of Euler
derivatives and the structural index. The objective of Euler estimates and can identify the correct structural index.
deconvolution is to obtain estimates of the base level of Values of declination and inclination in synthetic tests
the data, horizontal and vertical positions of the source, were based on Chulliat et al. (2014).
assuming a tentative structural index, in most of the
cases. Euler deconvolution

The structural index, that is related to the nature of the Considering a discrete set of N observations of total-field
source, can only be integer (Reid et al., 2014; Reid and anomaly, Euler deconvolution (Reid et al., 1990) can be
Thurston, 2014), otherwise the anomaly decay with written as a linear system of equations given by:
distance would change in a discontinuous way as the hi h h h h h
distance source to observer changes. This was clearly xˆo  yˆ o i  zˆo i  nbˆ  xi i  yi i  zi i  nhi , (1)
shown by Ravat (1996) using a dipole and a circular thin x y z x y z
disk (arbitrarily shaped source).
where hi  h( xi , yi , zi ) is the ith observation of the total-
field anomaly at the coordinates ( xi , yi , zi ) ,  is the
The gradients are usually calculated using numerical
methods. According to Reid et al. (2014) horizontal
gradients can be calculated using splines or finite structural index related to the nature or geometry of the
differences, while vertical gradients normally require source. The parameters estimated are xˆo , yˆ o , zˆo , related
to the horizontal and vertical coordinates of the source,

Fifteenth International Congress of the Brazilian Geophysical Society

 W HAT TO EXPECT FROM EULER DECONVOLUTION ESTIMATES 2
_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

borders of the anomaly, estimates are weighted by the

and b̂ , a base level (i.e., a background value). Therefore, coordinates of the moving data window, as pointed by
solving this linear system of equations provides four Silva and Barbosa (2003). Regardless of the structural
parameters, assuming a tentative structural index. Euler index used, estimates outside the source (at the borders
deconvolution works thru the whole dataset using a of the anomaly) are the same. An average of horizontal
moving data window scheme. At each window, Euler estimates can be done in these plateaus to achieve the
deconvolution solves the linear system for the four correct horizontal estimated value (Li, 2003). This
unknowns. In the following, will be present two tests with procedure is valid also for depth and base level
single anomaly with derivatives calculated analytically, estimates.
one simulating a vertical incidence and the other one an
inclined incidence of the field.
Vertical incidence
We generated a synthetic monopole (structural index = 2)
located at xo = 14 km, yo = 25 km and zo = 1 km,
intensity of 1 A/m and magnetized by induction. The
magnetic anomaly was calculated in a grid of 240 points
in x-coordinate (north direction) and 200 y-coordinate
(east direction). The grid starts in x = -10 km and y = 5 km
and it is equally spaced at each 0.2 km. Observation level
at this test is simulated at z = 1 km and base level is
equal to zero. The anomaly was corrupted with
pseudorandom Gaussian noise with standard deviation of
0.01 nT and seed 69. The anomaly is in Figure 1a and the
derivatives in Figures 1b-1d.

Figure 2 – X-estimates (a) – (c) and y-estimates (d) – (f)

from Euler deconvolution using an 11 x 11 moving data
window assuming structural indices 1, 2 and 3,
respectively for both estimates. Estimates using the
correct structural index are (b) and (e).
Figure 3 (a) – (c) displays depth estimates assuming
indices 1, 2 and 3, respectively. These estimates are the
ones used by Thompson (1982), Reid et al. (1990) and
others to select the correct structural index based on the
Figure 1 – Magnetic anomaly generated by a monopole clustering of solutions. We can identify in Figure 3 b) that
vertically magnetized with base level equal to zero and its estimates over the source have a constant value and thus
analytical derivatives. A) Magnetic anomaly. B) X form a plateau. Figure 3 b) exhibits the tight clustering of
derivative. C) Y derivative. D) Z derivative. solutions that those authors seek to identify when depth
estimates relates to the correct structural index. Only
We apply Euler deconvolution using an 11 × 11 moving these estimates, using the correct structural index, have
data window and assuming structural indices 1, 2 and 3. correct depth estimates. Using wrong structural indices
We plotted all estimates using the procedure adopted by estimates do not form plateaus. In Figure 3 a) we can see
Silva and Barbosa (2003), the estimates are plotted a cavity over the anomaly; this means that those depth
against the x- and y-coordinates of the center of the estimates are lower than the correct ones. While in Figure
moving data window used in Euler deconvolution. 3 c) we can see a prominence over the source; this
Figure 2 (a) – (c) displays x-estimates and Figure 2 (d) – means that depth estimates at this case are
(f) displays y-estimates from indices 1, 2 and 3, overestimated. Regardless of the structural index used, at
respectively. Estimates from all indices exhibit plateaus the borders of the anomaly the estimates are close to the
over the anomaly location. These plateaus mean that on simulated acquisition level, as pointed by Silva and
that area all estimated parameters have the same value. Barbosa (2003). Depth estimates form ramps at the limit
However, they are completely flat only in Figures 2 b) and between the borders of the anomaly and the region under
e), these are estimates using SI=2, the correct one for the influence of the anomaly. Thus, the minimum variation
this source. Using other SI is also possible to see the of depth estimates over the influence zone of the source,
plateaus in Figures 2 a), c), d) and f); however, they have bounded by those ramps, can define the correct structural
a small amount of undesired values in the center. These index, for the case of single isolated source.
values differ a bit from the correct one but nothing that
interfere in the final evaluation of the positions. At the

Fifteenth International Congress of the Brazilian Geophysical Society

 FELIPE MELO AND VALERIA BARBOSA 3
_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

depending if the structural index used is bigger or lower

than the correct one. At the region outside the source, at
the borders of the anomaly, values of base level
estimates are the correct one, regardless of the structural
index used, in this case it is zero. Thus, the minimum
variation of base level estimates over the source identify
the correct structural index, for the case of single isolated
source.

Figure 3 – Depth estimates of Euler deconvolution using Figure 4 – Base level estimates of Euler deconvolution
an 11 x 11 moving data window assuming structural using an 11 x 11 moving data window assuming structural
indices (a) 1, (b) 2 and (c) 3. Depth estimates using the indices (a) 1, (b) 2 and (b) 3. Base level estimates using
correct structural index are constant over the source and the correct structural index are constant over the source
have the smallest variation. and have the smallest variation.

Figure 4 (a) – (c) displays base level estimates from Euler Inclined magnetization
deconvolution assuming structural indices 1, 2 and 3, At this test, we kept the same geometric parameters of
respectively. From these estimates is easy to find an the source and grid. Now, acquisition is at ground level
outstanding pattern in the region over the anomaly at and geomagnetic inclination and declination were
Figure 4 b). The estimates showed in Figure 4 b) are from changed. The simulated geomagnetic field has inclination
to structural index 2, the correct one for this source. The of -39º and declination of -22º, values close to the field at
values of the estimates at this plateau are the correct the city of Rio de Janeiro. The magnetic anomaly is in
ones. Let us recall that the base level present in our Figure 5 (a) and the derivatives in Figures 5 (b) – (d).
synthetic data is zero. Moreover, when we compare the
range of the estimates, at the color bar for example, is
easy to see that the correct estimates have a lowest
variation compared to other estimates. Base level
estimates of Figures 4 a) and c) are protuberant and
show an angular pattern, positively or negatively,

Fifteenth International Congress of the Brazilian Geophysical Society

 W HAT TO EXPECT FROM EULER DECONVOLUTION ESTIMATES 4
_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

the interface of the source influence zone and the borders

of the anomaly because of the inclination and declination
simulated. However, is easy to distinguish between the
use of correct and wrong structural index due to the
plateau associated to the use of correct structural index,
Figure 7 (b). At the borders of the anomaly, the estimated
values are close to zero because simulated acquisition is
at ground level. Again, the minimum variation of depth
estimates over that plateau is still valid to define the
correct structural index.

Figure 5 – Magnetic anomaly generated by a monopole,

simulated geomagnetic field has inclination of -39º and
declination of -22º, with base level equal to zero and its
analytical derivatives. A) Magnetic anomaly. B) X
derivative. C) Y derivative. D) Z derivative.
We apply Euler deconvolution using the same data
moving window size, 11 × 11 points, and assuming
structural indices 1, 2 and 3.
Figure 6 (a) – (c) displays x-estimates and Figure 6 (d) –
(f) displays y-estimates from indices 1, 2 and 3,
respectively. As for the vertical incidence case, estimates
from all indices exhibit plateaus over the source. At these
estimates, plateaus are bigger than for vertical incidence
case; this is due to the inclination and declination of the
simulated geomagnetic field. The same pattern of small
disturbances appears with estimates using wrong
structural indices. At the borders of the anomaly, the
same analysis is valid and estimates are weighted by
moving data window.

Figure 6 – X-estimates (a) – (c) and y-estimates (d) – (f) Figure 7 – Depth estimates of Euler deconvolution using
from Euler deconvolution using an 11 x 11 moving data an 11 x 11 moving data window assuming structural
window assuming structural indices 1, 2 and 3, indices (a) 1, (b) 2 and (c) 3. Depth estimates using the
respectively for both estimates. Estimates using the correct structural index are constant over the source and
correct structural index are (b) and (e). have the smallest variation.
Figure 7 (a) – (c) displays depth estimates from Euler Finally, Figure 8 (a) – (c) displays base level estimates
deconvolution assuming indices 1, 2 and 3, respectively. from Euler deconvolution are assuming structural indices
These estimates do not form those ramps (Figure 3), at 1, 2 and 3, respectively. These estimates follow the same

Fifteenth International Congress of the Brazilian Geophysical Society

 FELIPE MELO AND VALERIA BARBOSA 5
_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

pattern of the vertical incidence case, only using the these analyses in anomalies with different simulated
correct structural index a plateau of correct estimates are geomagnetic field influence and we extend the results for
present. At the borders of the anomaly influence area, all cases. Horizontal estimates have almost the same
base level estimates are correct regardless of the used value using or not the correct structural index. Over the
structural index. Besides the plateau at the correct source, there are small differences on these estimates
estimate in Figure 8 (b), comparing the amplitudes range that make no difference in the result. At the borders of the
via the color bar is easy to identify that these estimates anomaly, moving data window coordinates weights the
have a small variation. Thus, like for the vertical incidence estimates and they are correct, regardless of the
test, the minimum variation of base level estimates can structural index used. Over the source, depth and base
define the correct structural index. level estimates show an outstanding pattern forming
plateaus only when the correct structural index is used.
This pattern allows the identification of the correct
structural index. More than that, depth and base level
estimates are only true over the source if the correct
structural index is used. Variation of base level and depth
estimates using the correct structural index is smaller
than estimates using the wrong structural index, over the
source. At the borders of the anomaly, depth estimates
are weighted by acquisition height while base level
estimates are the correct base level, regardless of the
structural index used, for both cases. These analyses
were possible because we have run Euler deconvolution
on an anomaly generated by a single isolated synthetic
source with high signal-to-noise level. Its derivatives are
analytic so no problems related to numerical derivatives
influence the results. We can extend these analyses to
any source that is valid for Euler deconvolution, any
values of base level, acquisition height and any inclination
or declination of the source or geomagnetic field, as long
as there are no interfering sources.
Acknowledgments
F. Melo was supported in this research by a Phd
scholarship from Coordenação de aperfeiçoamento de
Pessoal de Nível Superior (CAPES), Brazil. The first
author would like to thanks Bruno K., Francois D. and Roy
M. for grammatical suggestions.
References
Barbosa, C.F., Silva, B.C., Medeiros, W.E., 1999. Stability
analysis and improvement of structural index
estimation in Euler deconvolution. Geophysics 64,
48–60. doi:10.1190/1.1444529
Blakely, R.J., 1996. Potential Theory in Gravity and
Magnetic Applications. Cambridge University Press.
Chulliat, A., Macmillan, S., Alken, P., Beggan, C., Nair,
Figure 8 – Base level estimates of Euler deconvolution M., Hamilton, B., Woods, A., Ridley, V., Maus, S.,
using an 11 x 11 moving data window assuming structural Thomson, A., 2014. The US/UK World Magnetic
indices (a) 1, (b) 2 and (b) 3. Base level estimates using Model for 2015-2020. doi:10.7289/V5TH8JNW
the correct structural index are constant over the source Leao, J.W.D., Silva, J.B.C., 1989. Discrete linear
and have the smallest variation. transformations of potential field data. Geophysics
Conclusions 54, 497–507.
We have shown what to expect from Euler deconvolution Li, X., 2003. On the use of different methods for
estimates isolated sources. With this analysis was estimating magnetic depth. Lead. Edge 22, 1090–
possible to define that the minimum variation of either 1099. doi:10.1190/1.1634912
depth or base level estimates at the region of the source
can define the correct structural index. No further Melo, F.F., Barbosa, V.C.F., Uieda, L., Oliveira Jr, V.C.,
calculation is necessary to define the correct structural Silva, J.B.C., 2013. Estimating the nature and the
index. The average of the estimates over the plateau horizontal and vertical positions of 3D magnetic
associated with the correct structural index gives the sources using Euler deconvolution. Geophysics 78,
correct estimated value for any estimates. We performed J87–J98. doi:10.1190/GEO2012-0515.1

Fifteenth International Congress of the Brazilian Geophysical Society

 W HAT TO EXPECT FROM EULER DECONVOLUTION ESTIMATES 6
_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

Pašteka, R., Richter, F.P., Karcol, R., Brazda, K., Hajach,

M., 2009. Regularized derivatives of potential fields
and their role in semi-automated interpretation
methods. Geophys. Prospect. 57, 507–516.
doi:10.1111/j.1365-2478.2008.00780.x
Ravat, D., 1996. Analysis of the Euler method and its
applicability in environmental magnetic
investigations. J. Environ. Eng. Geophys. 1, 229–
238. doi:10.4133/JEEG1.3.229
Reid, A.B., Allsop, J.M., Granser, H., Millett, A.J.,
Somerton, I.W., 1990. Magnetic interpretation in
three dimensions using Euler deconvolution.
Geophysics 55, 80–91. doi:10.1190/1.1442774
Reid, A.B., Ebbing, J., Webb, S.J., 2014. Avoidable Euler
Errors - the use and abuse of Euler deconvolution
applied to potential fields. Geophys. Prospect. 62,
1162–1168. doi:10.1111/1365-2478.12119
Reid, A.B., Thurston, J.B., 2014. The structural index in
gravity and magnetic interpretation: Errors, uses,
and abuses. GEOPHYSICS 79, J61–J66.
doi:10.1190/geo2013-0235.1
Silva, J.B.C., Barbosa, V.C.F., 2003. 3D Euler
deconvolution: Theoretical basis for automatically
selecting good solutions. Geophysics 68, 1962–
1968. doi:10.1190/1.1635050
Telford, W.M., Geldart, L.P., Sheriff, R.E., 1990. Applied
Geophysics. Cambridge University Press,
Cambridge.
Thompson, D.T., 1982. EULDPH: A new technique for
making computer-assisted depth estimates from
magnetic data. Geophysics 47, 31.
doi:10.1190/1.1441278

Fifteenth International Congress of the Brazilian Geophysical Society