You are on page 1of 10

Journal of Applied Geophysics 162 (2019) 174–183

Contents lists available at ScienceDirect

Journal of Applied Geophysics

journal homepage: www.elsevier.com/locate/jappgeo

A comparative study for the source depth estimation of very low


frequency electromagnetic (VLF-EM) signals
A. Ebrahimi a,⁎, N. Sundararajan a,⁎, V. Ramesh Babu b
a
Department of Earth Science, Sultan Qaboos University, Muscat, Post Box: 36, Postal Code 123, Oman
b
Atomic Minerals Directorate for Exploration and Research, Hyderabad-500 016, India

a r t i c l e i n f o a b s t r a c t

Article history: In general, a quantitative interpretation of geophysical data yields information on the nature of subsurface
Received 7 June 2018 structures of geologic interest besides depth of the source and the associated physical property contrast. However,
Received in revised form 6 December 2018 all geophysical interpretation invariably incurs an inherent ambiguity including VLF-EM method. Techniques such
Accepted 10 January 2019
as Karous-Hjelt (K\\H) current density, Hilbert transform etc. are in vogue to estimate the depth to the source that
Available online 12 January 2019
are either qualitative or semi quantitative. Here, we present a comparative analysis of the depth derived from
Keywords:
VLF-EM signals by different methods over a uranium rich basement fractures of Raigarh District, Chhattisgarh,
In-phase component India. The obtained results are based on techniques that are not very common in VLF-EM data interpretation
Quadrature component such as Euler deconvolution (ED), Hartley spectral analysis and analytical signal approach of denoised in-phase
Depth estimation component realized by Empirical Ensemble Mode Decomposition (EEMD). The estimated depth from ED and
Euler deconvolution Hartley spectral analysis range 10–62 m and 12–40 m respectively are compared with K\\H pseudo section, Hilbert
Power spectrum transform and drilling depth. Overall, the results of the aforesaid techniques have shown satisfactory comparison
Empirical ensemble mode decomposition wherein the Hilbert transform of EEMD de-noised in-phase component of traverse T1 (35 m), Hartely power
(EEMD)
spectrum of the principal profile PPQ (40 m) and the radially averaged power spectrum (38.8 m and 40 m) are
close to the drilling depth. Therefore, the results obtained by methods presented emphasize that these techniques
are equally applicable to VLF-EM signals for estimation of depth to subsurface conductors.
© 2019 Elsevier B.V. All rights reserved.

1. Introduction such type of interpretation invariably incurs an inherent ambiguity


and therefore, we often employ multiple techniques to estimate the
Among all the geophysical methods for mineral prospecting, the depth to source from anomalies for more reliability. The measured
very low frequency electromagnetic (VLF-EM) method is seen as not in-phase (IP(x)) and out of phase (OP(x)) components commonly
only known for quick mapping of subsurface structures/conductors presented as stacked profiles as well as contour images. Generally, the
but also more reliable (Jang and Kim, 2015; Sasaki et al., 2015; Sharma interpretation of VLF-EM signal is based on qualitative analysis particu-
et al., 2014; Sundararajan et al., 2018). This method picks up signal radi- larly based on Fraser and K\\H filtering (Fraser, 1969; Karous and Hjelt,
ation from military navigation radio transmitters across the globe. There 1983) which are routinely used in many field including uranium explo-
are as many as 42 global ground military communication transmitters ration (Ramesh Babu et al., 2007), to decipher the nature of subsurface
operating in VLF frequency range of 15–30 kHz. The generated signals conductors besides their depth. In the recent past, Sundararajan et al.
are effectively used for many applications, including mineral explora- (2011) have used the method of Hilbert transform as a semi quantita-
tion (Sundararajan et al., 2006; Sharma et al., 2014; Al-Tarazi et al., tive analysis of IP(x) and OP(x) components for the estimation of
2008; Gürer et al., 2009). depth to source.
Estimation of depth to sources by electromagnetic methods plays an On the other hand, Euler deconvolution (ED) and spectral analysis,
important role in locating subsurface conductors and subsequent inter- which were widely used in potential field anomalies such as gravity,
pretation. A qualitative or quantitative interpretation of any geophysical magnetic, electric and electromagnetic fields to evaluate depth, are
data including VLF-EM signal basically focus structures of geologic inter- attempted here to interpret VLF-EM data. Initially, the ED was proposed
est, the depth and the relevant physical property contrast. However, by Thompson (1982) and later extended by Reid et al. (1990) and
improved by Keating (1998) and Mushayandebvu et al. (2004). ED
⁎ Corresponding authors.
has wide range of applications (Gerovska and Araúzo-Bravo, 2003)
E-mail addresses: p113935@squ.edu.om (A. Ebrahimi), visvid12@squ.edu.om and the popularity of this elegant method is due to its simplicity and
(N. Sundararajan). ease of implementation and hence its choice for a quick initial analysis.

https://doi.org/10.1016/j.jappgeo.2019.01.007
0926-9851/© 2019 Elsevier B.V. All rights reserved.
A. Ebrahimi et al. / Journal of Applied Geophysics 162 (2019) 174–183 175

Further, the radially averaged power spectral analysis has a wide range
of application in geophysical data processing and perhaps the most
common in potential field data for estimation of depth and other
parameters of the subsurface geological structures (Hinich and Clay,
1968; Bhattacharyya and Leu, 1977). In this direction, the spectral
analysis is implemented by an alternative tool called Hartley spectral
analysis (Sundararajan and Brahmam, 1998; Sundararajan et al., 2007)
that has a couple of advantages over traditional Fourier spectral analy-
sis, and it is being examined to VLF-EM data in this study.
In addition to these prevalent methods, reconstruction of noisy
VLF-EM signal with the empirical mode decomposition (EMD) and
ensemble empirical mode decomposition (EEMD) are relatively recent
methods and are examined here in this study. The process of EMD
separates a signal into several intrinsic mode functions (IMFs) which
is the residual of mean spline and import signal at each stage. An IMF
is nothing but a simple oscillatory mode with variable amplitude and
frequency in contrast with the simple harmonic possessing a constant
amplitude and frequency as functions of time/space. Further, certain
signals which cannot be separated to the same EMD stage every time
may be deciphered as mode mixing problem. In such case, an extended
process called ensemble empirical mode decomposition (EEMD) is
applied to separate signals into IMFs (Wu and Huang, 2005; Wu and
Huang, 2009; Wu et al., 2010)
The objective of applying EEMD technique to VLF-EM signals is to
decrease the effect of high frequency non-linear noise (Jeng et al.,
2007; Lin and Jeng, 2010; Sungkono Santosa et al., 2016). Jeng et al.,
(2007) applied EMD to VLF-EM data in order to reconstruct it from a
noisy raw VLF-EM signal. Lin and Jeng (2010) assessed the signal with
an empirical and qualitative technique to select optimum IMF compo-
nents, while Sungkono Santosa et al. (2016) defined a multivariate
Fig. 1. Stacked profile of VLF-EM IP(x) and OP(x) components namely T1 to T11, acquired
EMD for VLF-EM de-noising. The EEMD based reconstructed signal
in Raigarh District, India.

Fig. 2. Contour image of (a) IP(x) component with a principal profile PPI and (b) OP(x) component with principal profile PPQ.
176 A. Ebrahimi et al. / Journal of Applied Geophysics 162 (2019) 174–183

Fig. 3. ED indicating depth to source under the assumption of a dyke with structural index 1.0 from (a) IP(x) component (b) OP(x) component and a contact with structural index 0.5 from
(c) IP(x) component and (d) OP(x) component.
A. Ebrahimi et al. / Journal of Applied Geophysics 162 (2019) 174–183 177

may further be subjected to Hilbert transform analysis to determine a of the local geology and is given as (Stratton, 1941; Wait, 1982; Ward
more reliable depth as presented in this study. and Hohmann, 1988)
Here, a few well known depth estimation techniques that are in rffiffiffi
vogue in geophysical data interpretation are applied to a set of IP(x) and ρ
SD ≈ 500 ð1Þ
OP(x) components of VLF-EM data, from uranium rich basement frac- f
tures of Raigarh District, India. The results from all these methods are
compared and found to be close to each other in many cases. Thus, the where ρ is subsurface resistivity and f denotes the transmitter
results demonstrate that the efficiency and usefulness of these methods frequency. The depth derived by penetration of VLF-EM signal is varying
to the interpretation of VLF-EM data. from shallow to maximum of a few tens of m. However, it may be added
here that a conductive overburden will firmly damp signals while a
resistive environ is ideal for a VLF-EM survey.
2. Methodology Karous and Hjelt (K\\H) filter of IP (x) and OP(x) is a forward
modeling technique for depth estimation. In K\\H filtering, the
Let IP (x) and OP(x) be the in-phase and out of phase components of VLF-EM signal is filtered at a range of station interval (related to skin
VLF-EM signal, from which we derive the depth to source by different depth) between measuring points and can be defined as (Karous and
methods such as K\\H filtering, ED (Euler deconvolution), Hartley spec- Hjelt, 1983)
tral analysis and HT (Hilbert transform) of EEMD, which are briefly    
discussed in this section. Δz Δx
Ia ¼ −0:205:H−2 þ 0:323:H −1 −1:446:H 0
In general, the depth of investigation is defined by the skin depth 2π 2
(SD) in EM data interpretation that depends on the electrical properties þ 1:446:H 1 −0:323:H 2 þ 0:205:H3 ð2Þ

Fig. 4. Radially averaged power spectrum of VLF-EM (a) IP(x) and (b) OP(x) components. Hartley Power Spectrum of Principal Profiles (c) PPI and (d) PPQ.
178 A. Ebrahimi et al. / Journal of Applied Geophysics 162 (2019) 174–183

Fig. 5. Hartley Power Spectrum of IP(x) and OP(x) components of traverses (a and b) IP(%) and OP(%) of T1, (c and d) IP(%) and OP(%) of T6 and (e and f) IP(%) and OP(%) of T11.
A. Ebrahimi et al. / Journal of Applied Geophysics 162 (2019) 174–183 179

Fig. 6. Intrinsic mode function of traverse T1.

where Δz is assumed as thickness of the current sheet, Ia is the current The maximum amplitude of analytic signal in the plot of x versus
density, and Δx is the distance between the data points and also the AA(x) corresponds to the center of the subsurface source as in the
depth to the current sheet. case of Fraser filter peaks (Sundararajan et al., 2011) whereas the
In Hilbert transform analysis, VLF-EM signals IP (x) and OP(x) depth to the source can be inferred as the half sum of the abscissa of
are subjected to computing Hilbert transform and analytical signal the points of the intersection of IP(x) and HT(x) as in the case of gravity
expressed as: and magnetic data interpretation (Sundararajan and sirvings 2010;
Sundararajan et al., 2011). However, when the width of the target is
AðxÞ ¼ IPðxÞ þ i HTðxÞ ð3Þ larger, the function AA(x) gives rise to two peaks with a minimum at
the center. The minimum of AA(x) correspond to the center of the
where HT(x) is the Hilbert transform of IP(x) that can be computed target, and the distance between the two peaks yield the width of the
as discussed by Sundararajan and Srinivas (2010). The amplitude target (Nabighian, 1974).
AA(x) of the analytical signal can be obtained as Fourier spectral analysis of geophysical signals particularly the
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi potential field data are well known and common for the estimation of
AAðxÞ ¼ IPðxÞ2 þ HTðxÞ2 ð4Þ the source parameters including depth, width, dip etc. (Bhattacharyya
and Leu, 1977; Hinich and Clay, 1968). In this direction, the Hartley

Fig. 7. Intrinsic mode function of traverse T7.


180 A. Ebrahimi et al. / Journal of Applied Geophysics 162 (2019) 174–183

Fig. 8. Intrinsic mode function of traverse T11.

spectral analysis is a real replacement of complex Fourier analysis and is original signal by which form, a complete and nearly orthogonal filter-
often used in the recent past for similar studies (Sundararajan et al., ing functions. The resulting functions, known as intrinsic mode func-
2007). As in the Fourier spectral analysis, the slope of the log power tions (IMFs), are therefore sufficient to describe the signal, in spite of
spectrum of the plot of IP(x) versus wavenumber, yields the required the fact that they are non-orthogonal. Therefore, the IP(x) component
depth to the source. can be written as
Further, the ED is a widely used technique in deriving the depth to
source in many cases of potential field particularly magnetic data on X
n
IP ðxÞ ¼ Ci þ sn ð6Þ
the horizontal location of an anomalous source. In this case, the ED re- i¼1
quires spatial location of the points and a parameter called the structural
index (SI). The 3-D Euler's equation for IP(x) in the case of VLF-EM signal where Ci (i = 1, …, n) are components of IMFs and sn is the residue
can be expressed some what similar to Thompson (1982) as (Huang et al., 1998). Further details on EMD are available in literature
(Huang and Wu, 2008).
∂IP ðxÞ ∂IP ðxÞ ∂IP ðxÞ A staggering signal by EMD processing usually results in mode
ðx−x0 Þ þ ðy−y0 Þ þ ðz−z0 Þ ¼ N IP ðxÞ ð5Þ
∂x ∂y ∂z mixing problem and cause the decomposition to be unstable (Wu
et al., 2010). In order to overcome this, an extended technique of en-
where z is depth, (x, y) are observation coordinates, (x0, y0, z0) corre- semble empirical mode decomposition (EEMD) is being used (Wu and
sponds to source position , and N is positive that is related to the homo- Huang, 2005). The EEMD is working on the basis of adding a white
geneity of the potential and also sometimes regarded as structural index noise series and evaluating the new IMF components and the rest of
(SI) (Stavrev and Reid, 2007), it differs from field to field and types of the process remains the same as EMD. The added white noise aids in
source. gathering a signal in time-frequency domain uniformly with the
Empirical mode decomposition (EMD) is a method that aids in resulting components of variant scales separated by the EEMD function
decomposing a signal without leaving the time/space. For all natural sig- (Wu and Huang, 2009). In analyzing the IMF components, the first com-
nals that are non-stationary and non- linear, this technique is extremely ponent of IMFs includes high frequency noise which is a common phe-
useful for a reliable analysis. In this case, EMD acts as a basis for the nomenon in VLF-EM signal and the last one is just the residue (Lin and

Fig. 9. The EEMD filtered signal of T1, the Hilbert transform and the amplitude of the analytic signal.
A. Ebrahimi et al. / Journal of Applied Geophysics 162 (2019) 174–183 181

Fig. 10. The EEMD filtered signal of T7, the Hilbert transform and the amplitude of the analytic signal.

Fig. 11. The EEMD filtered signal of T11, the Hilbert transform and the amplitude of the analytic signal.

Table 1
The depth derived from Hartley Power spectrum and ED of IP(x) and OP(x) components of
VLF-EM signals.
Table 2
VLF-EM Signal Hartley Power spectrum ED The depth derived with Hilbert transform of EEMD-filtered IP(x) component of VLF-EM
signal.
IP (%) OP (%) IP (%) OP (%)
The depth derived from Hilbert h width(w1) width(w2)
T1 12.8 m 15.6 m 11-54 m 13.5-62 m
transform of EEMD-filtered Signal
T6 11.4 m 11.2 m
T 11 11.6 m 12 m T1 35 m – –
Principal profiles 31.5 m (PPI) 40 m (PPQ) T7 8–18 m 50 m 182 m
Radially averaged 38.8 m 40 m T11 20 m 128 m –
182 A. Ebrahimi et al. / Journal of Applied Geophysics 162 (2019) 174–183

Jeng, 2010; Huang et al., 1998). De-noising VLF-EM signal by EMD and

(Ramesh Babu
Drilling depth

et al., 2007)
EEMD techniques, in a few recent publications, Jeng et al. (2007), Lin
and Jeng (2010) and Sungkono Santosa et al. (2016) used a combination

36–145
of other components to reconstruct VLF-EM signal.
The Hilbert transform of an EEMD de-noised VLF-EM signal can be
considered further to estimate the depth to subsurface source. By

ED of total magnetic

Babu et al., 2007)


using the IP(x) or OP(x) component along with Hilbert transform and

field (Ramesh
amplitude of analytical signal (Eq. (4)) one can estimate the depth to
top of the conductor (Sundararajan et al., 2011 and Ndatuwong and
Yadav, 2013).

15–40
3. Data

current density
pseudo section

10–50
(Ramesh Babu
Fraser filtered

et al., 2007)
To test the applicability of various interpretation methods as



discussed in above section that are not common to VLF-EM data analysis

IP (%)
to the available VLF-EM data is carried out in this study. The data con-

T11
T7
T5
sists of 11 traverses namely T1 to T11 of IP(x) and OP(x) components
expressed in percentile that were acquired in uranium rich basement

section (Ramesh
density pseudo

10–45

10–40
fractures in Sambalpur granitoids of Raigarh district, Chhattisgarh,

et al., 2007)
K-H current
India (Sundararajan et al., 2006 and Ramesh Babu et al., 2007). In the


area of study, the resistivity of host rock is found to be 800 Ω-m

IP (%)
Babu

T11
(Ramesh Babu et al., 2007). The data were recorded at an interval of

T7
T5
10 m at a transmitter frequency 22.2 kHz corresponding to JJI Ebino,
Japan located at latitude 32.092247 N and longitude 130.829095 E.

(Sundararajan

35

30
The IP(x) and OP(x) components of the traverses are given in the

et al., 2011)


transform
form of stacked profiles and shown in Fig. 1. On the other hand, Fig. 2

Hilbert

IP (%)
(a) and (b) show the contour image of IP(x) and OP(x) with principal

T11
T7
T5

PP
profiles namely PPI and PPQ. As discussed in methodology, the ED


of IP(x) and OP(x) components were computed and shown in Fig. 3

50–182
Signal (Present
EEMD-filtered
(a–d). Further, the radially averaged power spectrum of IP(x) and OP

8–18
transform of

128
35

20
(x) were computed in addition to the power spectrum of the principal

method)
Hilbert
profiles PPI and PPQ of IP(X) and OP(x) and shown in Fig. 4(a–d). The

IP (%)

T11

T11
T7
power spectrum of IP(x) and OP(x) of a few representative traverses

T1

T7
namely T1, T6 and T11 are illustrated in Fig. 5(a–f). The intrinsic mode
function (IMF) of the IP(x) component of traverses T1, T7 and T11
OP(%)

(PPQ)
15.6
were computed based on the application of ensemble empirical mode

13
12
40

40
decomposition (EEMD) and shown in Figs. 6, 7 and 8. The results of

31.5 (PPI)
IMFs namely IMF1, IMF2 and IMF3 represent various levels of noise
whereas IMF4 (in the case of T1 and T7) indicates the residual part
12.8
12.5
11.6

38.8
with least noise, however, is not used for further analysis. In the case
Hartley Power spectrum

of traverse T11, IMF3 is the residual. In all the other cases, IMF1 is too
Principal Profile (PP)

noisy to be considered for the reconstruction of original signal. And


(Present method)

Radially averaged

therefore, the sum of IMF2 and IMF3 (in the case of T1 and T7) are con-
Comparison of depths (in meter) derived from VLF-EM signals by different techniques.

sidered as reconstructed signal and in the case of T11, IMF2 itself is used
as reconstructed signal. Then, the reconstructed signals are used for Hil-
IP (%)

bert transform analysis to derive the depth to the source as a function of


T11
T6
T1

abscise of the points of intersection of in-phase (IP(x) and HT(x))


(Sundararajan and Srinivas, 2010) and are shown in Figs. 9, 10 and 11,
– 11–32 (contact)

along with amplitude of analytic signal that is used for precise spatial lo-
13.5–62 (dyke)

cation of subsurface target.


ED of VLF-EM (Present method)

OP(%)

4. Results and discussion

The depth to the subsurface source were obtained by different


– 10–41 (contact)

methods such as Hartley spectral analysis, ED and Hilbert transform


11–54 (dyke)

analysis of ensemble empirical mode decomposition (EEMD) of IP


(x) component (Table 1 and Table 2). The role of EEMD was to improve
IP(%)

the quality of noisy IP(x) component. Thus, the depth obtained from all
these methods were compared in turn with the available drilling depth

as well as the depth obtained from magnetic data (Ramesh Babu et al.,
Width of source (m)

2007). Initially, the ED of IP(x) and OP(x) components have resulted


the depth to source under an assumption of a dyke (structural index
Parameters

1.0) in the range 11–54 m and 13.5–62 m and a contact (structural


Depth (m)

index 0.5) in the range 10- 41 m and 11–32 m respectively (Fig. 3).
Table 3

On the other hand, the radially averaged power spectral analysis of


both IP(x) and OP(x) yield depth that range 39 m and 40 m (Fig. 4
A. Ebrahimi et al. / Journal of Applied Geophysics 162 (2019) 174–183 183

(a) and (b)) respectively, however, the spectral analysis of the principal Fraser, D., 1969. Contouring of VLF-EM data. Geophys. 34 (6), 958–967.
Gerovska, D., Araúzo-Bravo, M.J., 2003. Automatic interpretation of magnetic data based
profiles PPI and PPQ from IP(x) and OP(x) resulted depth around 31.5 m on Euler deconvolution with unprescribed structural index. Comput. Geosci. 29 (8),
and 40 m respectively (Fig. 4(c) and (d)). Out of the eleven traverses, 949–960.
the Hartley spectral analysis of the first (T1), middle (T6) and the last Gürer, A., Bayrak, M., Gürer, Ö.F., 2009. A VLF survey using current gathering phenomena
for tracing buried faults of Fethiye–Burdur Fault Zone, Turkey. J. Appl. Geophys. 68
traverses (T11) yield depths in the range 12.8–15.6 m, 11.2–11.4 m (3), 437–447.
and 11.6–12 m (Fig. 5 and Table 1) respectively. Thus, all the depths par- Hinich, M.J., Clay, C.S., 1968. The application of the discrete fourier transform in the esti-
ticularly from spectral analysis of traverses almost lie in the same range. mation of power spectra, coherence, and bispectra of geophysical data. Rev. Geophys.
6 (3), 347–363.
Further, the depths from the Hilbert transform analysis of EEMD based Huang, N.E., Shen, Z., Long, S.R., Wu, M.C., Shih, H.H., Zheng, Q., Liu, H.H., 1998. The empir-
IP(x) component of traverses T1, T7 and T11 are found to be 35 m, 8– ical mode decomposition and the hilbert spectrum for nonlinear and non-stationary
18 m and 20 m respectively (Figs. 9, 10 and 11 and Table 2). In addition, time series analysis. in proceedings of the royal society of london A: mathematical.
Phys. Eng. Sci. 454 (1971), 903–995.
the HT analysis have also resulted the width of the subsurface target
Huang, N.E., Wu, Z., 2008. A review on hilbert huang transform: method and its applica-
(the distance between the peaks of the amplitude of analytical signal) tions to geophysical studies. Rev. Geophys. 46 (2).
as 50–182 m and 128 m for traverses T7 and T11 respectively. It may Jang, H., Kim, H.J., 2015. Mapping deep-sea hydrothermal deposits with an in-loop tran-
be noted that the amplitude of analytical signal not only ensures precise sient electromagnetic method: Insights from 1D forward and inverse modeling.
J. Appl. Geophys. 123, 170–176.
spatial location of subsurface targets, but also the width of the target, Jeng, Y., Lin, M.J., Chen, C.S., Wang, Y.H., 2007. Noise reduction and data recovery for a VLF-
when the depth is less than the width (Nabighian, 1974). Further, the EM survey using a nonlinear decomposition method. Geophys. 72 (5), F223–F235.
depth from drilling is seen to be comparable in many cases and agrees Karous, M., Hjelt, S.E., 1983. Linear filtering of VLF dip-angle measurements. Geophys.
Prospect. 31 (5), 782–794.
well with the interpreted depths by different methods (Table 3). Keating, P.B., 1998. Weighted euler deconvolution of gravity data. Geophysics 63,
It may be ascertained that the EMD is an adaptive signal decomposi- 1595–1603.
tion algorithm for the analysis of non-stationary and nonlinear signals Lin, M.J., Jeng, Y., 2010. Application of the VLF-EM method with EEMD to the study of a
mud volcano in southern taiwan. Geomorphology. 119 (1), 97–110.
(i.e. signals generated from nonlinear systems), and therefore it is Mushayandebvu, M.F., Lesur, V., Reid, A.B., Fairhead, J.D., 2004. Grid Euler deconvolution
gaining great importance in the recent past in almost all fields including with constraints for 2D structures. Geophys. 69 (2), 489–496.
geophysical studies. It decomposes the given data into a set of finite Nabighian, M.N., 1974. Additional comments on the analytic signal of two-dimensional
magnetic bodies with polygonal cross-section. Geophys. 39 (1), 85–92.
number of narrow band intrinsic mode functions (IMFs) which are Ndatuwong, L.G., Yadav, G.S., 2013. Analysis and interpretation of in-phase component of
derived directly from the data, unlike other signal decomposition tech- VLF-EM Data using hilbert transform and the amplitude of analytical signal. J. Env. Sci.
niques such as Fourier and Wavelet transforms incorporate predefined 3 (11).
Ramesh Babu, V., Ram, S., Sundararajan, N., 2007. Modeling and inversion of magnetic and
fixed basis for signal modeling and analysis. The ensemble EMD
VLF-EM data with an application to basement fractures: a case study from Raigarh,
(EEMD) is a noise-assisted data analysis method is considered to over- india. Geophys. 72 (5), B133–B140.
come the time-scale separation problem of EMD. Although a consider- Reid, A.B., Allsop, J.M., Granser, H., Millett, A.T., Somerton, I.W., 1990. Magnetic interpreta-
able success, all of the EMD algorithms are based on empirical, tion in three dimensions using euler deconvolution. Geophys. 55 (1), 80–91.
Sasaki, Y., Yi, M.J., Choi, J., Son, J.S., 2015. Frequency and time domain three-dimensional
heuristic and ad hoc procedures that make them hard to analyze inversion of electromagnetic data for a grounded-wire source. J. Appl. Geophys.
mathematically, and EMD may suffer from mode mixing, de-trend un- 112, 106–114.
certainty, aliasing and end effect artefacts (Singh et al., 2017). Thus, Sharma, S.P., Biswas, A., Baranwal, V.C., 2014. Very Low-Frequency Electromagnetic
Method: A Shallow Subsurface Investigation Technique for Geophysical applications.
EMD/EEMD are user friendly algorithms in applications to derive certain In Recent Trends in Modelling of Environmental Contaminants. Springer India.
desired outcome despite the fact that it is yet to establish firm mathe- pp. 119–141.
matical foundation. Singh, P., Joshi, S.D., Patney, R.K., Saha, K., 2017. The Fourier decomposition method for
nonlinear and non-stationary time series analysis. Proc. R. Soc. A. 473 (2199),
20160871.
5. Conclusions Stavrev, P.Y., Reid, A.B., 2007. Degrees of homogeneity of potential fields and structural in-
dices of euler deconvolution. Geophys. 72, L1–L12.
Stratton, J.A., 1941. Electromagnetic Theory. McGraw-Hill Book Co., New York.
In general, the interpretation of VLF-EM signal is qualitative in nature Sundararajan, N., Brahmam, G.R., 1998. Spectral analysis of gravity anomalies caused by
and is being confined to Fraser and K\\H filtering. The methods that are slab-like structures: a hartley transform technique. J. Appl. Geophys. 39 (1), 53–61.
not common to VLF-EM interpretation such as ED, power spectral anal- Sundararajan, N., Ramesh Babu, V., Shiva Prasad, N., Srinivas, Y., 2006. Short note:
VLFPROS-A matlab code for processing of VLF-EM data. Comput. Geosci. 32 (10),
ysis and EEMD based Hilbert transform techniques in this study, result 1806–1813.
depths to source that are comparable to the ones obtained from the tra- Sundararajan, N., Al-Garni, M.A., Ramabrahmam, G., Srinivas, Y., 2007. A real spectral
ditional techniques. The results that are based on power spectral analy- analysis of the deformation of a homogenous electric field over a thin bed–A Hartley
transform approach. Geophys. Prospect. 55 (6), 901–910.
sis of traverses as well as the principal profiles remain almost consistent Sundararajan, N., Srinivas, Y., 2010. Fourier–hilbert versus hartley–hilbert transforms
all through and are close to depths derived from Hilbert transform. The with some geophysical applications. J. Appl. Geophys. 71 (4), 157–161.
actual depths to source obtained from drilling also are comparable Sundararajan, N., Babu, V.R., Chaturvedi, A.K., 2011. Detection of basement fractures
favourable to uranium mineralization from VLF-EM signals. J. Geophys. Eng. 8 (2),
with depths derived from different methods, and thus substantiating 330.
the present analysis that these methods can also be practical in VLF- Sundararajan, N., Pracejus, B., Al-Khirbash, S., Al-Hosni, T.K., Ebrahimi, A., Al-Mashani, M.,
EM data interpretation as applicable in other geophysical data. Al-Lazki, A., 2018. Geophysical and geochemical investigation for structures favorable
to uranium mineralization in dhofar region, sultanate of oman. Arab. J. Geosci. 11
(22), 700.
Acknowledgment Sungkono Santosa, B.J., Bahri, A.S., Santos, F.M., Iswahyudi, A., 2016. Application of noise-
assisted multivariate empirical mode decomposition in VLF-EM Data to identify
underground river. Adv. Data Sci. Adapt. Anal. 1650011.
The authors would like to thank the reviewers particularly Prof. Thompson, D.T., 1982. EULDPH: a new technique for making computer-assisted depth es-
Jianghai Xia for providing a constructive and useful review of the man- timates from magnetic data. Geophys., vol 47, 31–37.
uscript that helped in the revision as presented. Dr V. Ramesh Babu, Wait, J.R., 1982. Geo-Electromagnetism. Academic Press, New York.
Ward, S.H., Hohmann, G.W., 1988. Electromagnetic theory for geophysical applications.
Atomic Minerals Directorate for Exploration and Research, Hyderabad, In: Nabighian, M.N. (Ed.), Electromagnetic Methods in Applied Geophysics, Theory
India, is acknowledged for his helpful digital data. —vol. 1. SEG, Tulsa, Oklahoma, pp. 131–308.
Wu, Z., Huang, N.E., 2005. Statistical Significance Test of Intrinsic Mode Functions.
Hilbert–Huang Transform and Its Applications. 5 pp. 107–127.
References
Wu, Z., Huang, N.E., 2009. Ensemble empirical mode decomposition: a noise-assisted data
analysis method. Adv. Adapt. Data Anal. 1 (01), 1–41.
Al-Tarazi, E., Rajab, J.A., Al-Naqa, A., El-Waheidi, M., 2008. Detecting leachate plumes and
Wu, S.D., Chiou, J.C., Goldman, E., 2010. Solution for mode mixing phenomenon of the em-
groundwater pollution at Ruseifa municipal landfill utilizing VLF-EM method. J. Appl.
pirical mode decomposition. In 2010 3rd International Conference on Advanced
Geophys. 65 (3–4), 121–131.
Computer Theory and Engineering (ICACTE). 2 pp. V2–500 (IEEE).
Bhattacharyya, B.K., Leu, L.K., 1977. Spectral analysis of gravity and magnetic anomalies
due to rectangular prismatic bodies. Geophys. 42 (1), 41–50.

You might also like