Journal of Applied Geophysics 153 (2018) 47–63

Contents lists available at ScienceDirect

Journal of Applied Geophysics

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

Exploration of buried carbonate aquifers by the inverse and forward

modelling of the Controlled Source Audio-Magnetotelluric data
Franjo Šumanovac ⁎, Jasna Orešković
University of Zagreb, Faculty of Mining, Geology and Petroleum Engineering, Pierottijeva 6, 10000 Zagreb, Croatia

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

Article history: On the selected cases, Gotalovec in the area of Pannonian basin and Baška in the Dinaridic karst area, that are
Received 19 October 2017 representing a common hydrogeological model in both regions of Croatia, CSAMT data together with data of
Received in revised form 28 March 2018 other geophysical methods (electrical resistivity tomography, electrical sounding and seismic reflection) enabled
Accepted 7 April 2018
the definition of a reliable prognostic geological model. The model consists of carbonate aquifer which underlies
Available online 11 April 2018
an impermeable thick package of clastic deposits. There are great variations of the dolomitic aquifer depths in the
Keywords:
Gotalovec area due to strong tectonic activity, while in the Baška area depth changes are caused by the layer
CSAMT folding.
Groundwater The CSAMT method provides the most complete data on lithological and structural relationships in cases of
Carbonate aquifer hydrogeological targets deeper than 100 m. Based on the presented models we can conclude that the CSAMT
Forward modelling method can provide greater exploration depth than electrical resistivity tomography (ERT) and can be consid-
Inversion ered as a fundamental geophysical method for exploration of buried carbonate aquifers, deeper than 100 m.
But, the CSAMT research may demonstrate its advantages only in the case of very dense layout of CSAMT stations
(25–50 m), due to the greater sensitivity to noise in relation to resistivity methods. Interpretation of CSAMT data
is more complex in relation to resistivity methods, and a forward modelling method sometimes gives better re-
sults than an inversion due to possibility of the use of additional data acquired by other geophysical methods
(ERT, electrical sounding and seismic reflection). At greater depths, the resolution of all electrical methods includ-
ing the CSAMT method is significantly reduced, and seismic reflection can be very useful to resolve deeper lith-
ological interfaces.
© 2018 Elsevier B.V. All rights reserved.

1. Introduction On the other hand, electrical resistivity tomography (ERT) can be

Magnetotelluric method (MT-method) has been applied for a long
time for deep geological investigations, mainly geothermal and hydro-
carbon investigations at depths of up to 10 km (e.g. Vozoff, 1972;
Volpi et al., 2003; Spichak and Manzella, 2009), but also in very deep in-
vestigations of the Earth's crust and the lithospheric mantle which ap-
plies to depths up to 100 km (e.g. Pous et al., 1995; Unsworth et al.,
2004; Semenov et al., 2008). It has only been used in recent decades
in relatively shallow investigations in engineering geology and hydro-
geology studies by the development of the method version which can
use high frequency range responsible for near surface targets. These
methods have been called Controlled Source Audio-Magnetotellurics
(CSAMT) and Controlled Source Radio Magnetotellurics (CSRT). The
methods can contribute to better determination of near surface geolog-
ical relationships (Ismail et al., 2015; Pedrera et al., 2016).
⁎ Corresponding author at: University of Zagreb, Faculty of Mining, Geology and
Petroleum Engineering, Pierottijeva 6, HR-10.000 Zagreb, Croatia.
E-mail address: franjo.sumanovac@rgn.hr. (F. Šumanovac).
considered as the most recent method in near surface geophysical ex-
plorations (Dahlin, 1996; Loke and Barker, 1995), and its introduction
has significantly advanced geophysical research. In the beginning the
method was applied for shallow engineering geology investigations,
but soon after it was applied in hydrogeological investigations, up to a
depth of 100 m. Results of the first applications showed that the ERT
can provide for a reliable predicted geological model, especially in the
case of complex geological models, where the geological models of
karstic terrains certainly belongs (Šumanovac and Weisser, 2001;
Šumanovac et al., 2003). Besides, the karstic terrains are characterized
by difficult conditions of field measurement. Results of many investiga-
tions in the karstic region of Croatia in the framework of groundwater
research forced the conclusion that ERT has been considered to be the
fundamental method in karstic hydrogeological investigations (Terzić
et al., 2007; Vlahović and Šumanovac, 2015). The other geophysical
methods, especially the seismic ones, are used in order to solve specific
tasks and reduce the ambiguity of geophysical interpretation
(Engelsfeld et al., 2008, 2011). The main reason for such application
lies in the fact that resistivity methods involve the entire volume of

https://doi.org/10.1016/j.jappgeo.2018.04.007
0926-9851/© 2018 Elsevier B.V. All rights reserved.
48 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63
rock; based on the measured rock resistivity, and data about rock type
and rock condition (whether it is compact, fractured or porous) can be
obtained.
Electrical tomography is also applied to solve geological models in
sedimentary basin. However, in the case of aquifers composed of thin
sandy or gravelly layers, the other geophysical methods should be ap-
plied, especially seismic reflection, to determine a reliable predicted
geological model (Šumanovac, 2006).
Standard equipment for ERT is efficient up to depths of about 100 m,
but in the case of deeper targets in hydrogeological and geothermal in-
vestigations, ERT is not usually carried out because the field equipment
would be too expensive and impractical. The tomographic profiles
should be very long to reach deep targets, and there is often not enough
space to set such long profiles. Length of the profiles could be often 800
m only, which means the depth penetration of 130 m in the middle of
the profile.
Deeper hydrogeological investigations seek depth penetration of
several hundred meters, and geothermal investigations even more
than 1 km. In such cases, standard ERT cannot be applied in an econom-
ical and routine way, so other geophysical methods should be consid-
ered. The use of CSAMT can provide greater depths of research, as well
as good space sampling and relatively dense measurements. A common
hydrogeological model, which can be found in both main regions in
Croatia, Pannonian basin and Dinarides (Fig. 1), can be solved by the
CSAMT method. The carbonate aquifer is underlying an impermeable
package of clastic deposits. The main goal is to define as precisely as pos-
sible the carbonate-clastic interface which will enable determination of
optimal exploratory borehole positions. Two characteristic examples of
groundwater investigations carried out in the Pannonian basin and
Dinarides in Croatia are shown. The results achieved by the CSAMT
method were correlated with data obtained by other geophysical
methods, seismic reflection and electrical resistivity tomography, for
Fig. 1. Positional map of the surveys. Locations 1 and 2 belong to Pannonian basin, while location 3 belongs to the Dinarides. 1 – position of the profile P-1 at Gotalovec, situated in a hilly
area; 2 – position of the profile P-2 at Gotalovec, situated in a valley; 3 – position of the profile P-3 on karstic island Krk.
comparison and verification. The main disadvantage of the CSAMT
method compared to the direct current methods is greater sensitivity
to urban and surface noises, which is especially prominent in urban en-
vironments and in the vicinity of electrical infrastructural facilities.
2. Geological setting
The targets of geophysical explorations in Croatia are often carbon-
ate rocks that underlie clastic deposits. Carbonate rocks, Triassic dolo-
mites and Jurassic and Cretaceous limestones and dolomites, may be
tectonically fractured and weathered and are therefore permeable and
as such act as aquifer. Clastic sediments above them consist mainly of
impermeable clay, shale and marl, which are insulating rocks. This
model is found in both geological regions in Croatia, in the Pannonian
Basin and the Dinaridic karst area (Fig. 1). In the Dinarides clastic de-
posits are usually Eocene flysch sediments. Although they may contain
permeable rocks such as sands, sandstones and carbonates, they are in
a package of impermeable rocks, clay and marl, so the entire package
acts as an impermeable barrier. Two representative locations are se-
lected in order to analyse the possibilities of CSAMT method in ground-
water research: Gotalovec in the Pannonian Basin and Baška on island
Krk that belongs to the External Dinarides.
2.1. Gotalovec area
The first study area - Gotalovec, is located in the north-western
Croatia, at the foot of Mount Ivančica (Fig. 1). The geological model in
this part of Croatia consists of Quaternary and Neogene sediments
(Miocene layers), and Cretaceous volcanoclastic rocks (K1,2) which dis-
cordantly lay on a surface made of Upper and Middle Triassic dolomites
(T3, T2), Šimunić et al. (1982; Fig. 2). These dolomites were exposed to
intense tectonic activities in several stages during geological history.
 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63
For that reason the secondary, fracture porosity was created and the
ground rock massif of the dolomites have become the main aquifer in
this part of Croatia with a significant water supply potential. Neogene
rocks (Miocene and Pliocene) are mostly built of marls, clays, shales
and fine-grained sandstones are impermeable as such. Occasionally
Lithothamnium limestone from Badenian can be found in the middle
part of Miocene deposits, at the contact with Triassic dolomites. They
come as irregularly-shaped bodies, in reef development, and are also
aquifers when in contact with dolomites. The Lower Triassic clastic sed-
iments are placed beneath the Triassic carbonate aquifer, which are con-
sidered as impermeable or low permeable layers. In view of the fact that
impermeable clastics and permeable carbonates are clearly distin-
guished by their resistivities, this model can be very efficiently investi-
gated using electrical and electromagnetic methods.
Low mineral and thermal water (temperature of 25.6 °C) from the
spring Topličica, close to the study area (Fig. 2), according to the ionic
composition belongs to the CaMg-HCO–3 (calcium magnesium hydrogen
carbonate) mixed type (Borović et al., 2016). This water type also points
to the aquifer which is built of Badenian limestone and Triassic dolo-
mites as a unique aquifer. The water has been recently utilized for
water bottling, as well. The Triassic dolomites appear at the surface on
Mount Ivančica, which are also connected to the dolomite basement
in the Mountain foothills enabling water recovery in the aquifer.
Dragičević et al. (1998) evaluated secondary porosity in the range 5–
25% in the mountain aquifers based on observing outcrops on larger
quarries. The aquifer thicknesses vary considerably ranging from 200
m to 1800 m. They also concluded on the basis of established long pe-
riod balance between recharge and storage that renewable resources
for these aquifers are similar to the annual discharge at aquifer springs.
According to previous investigations conducted by an exploratory
borehole located at one of the deep electrical sounding site in the
wider area (Fig. 2), Pliocene and Miocene clastics, mostly marls, are
Fig. 2. Geological map (Šimunić et al., 1982) and positions of the profiles at the Gotalovec area placed in the Pannonian basin.
49
found between the surface and a depth of 266.5 m, followed by Mio-
cene, Upper Badenian Lithothamnium limestone to a depth of 339 m,
and Triassic dolomites (Pickel, 2004). Lithothamnium limestone ap-
pears in the form of reef, and thus large lateral changes in its thickness
can be expected.
2.2. Baška area
The second micro-location is in the south-eastern part of Adriatic is-
land Krk, which belongs to the External Dinarides and can be character-
ized as a karstic area (Fig. 1). According to the previous geological
investigations Eocene flysch layers (E2,3), Eocene limestones (E1) and
Upper Cretaceous carbonates (K22, 2K1,2 2 ) are found at the survey area
(Mamužić et al., 1969; Fig. 3). Eocene flysch layers are mostly composed
of a package of clastic deposits (marls and sandstones) with limestone
inserts. Although the package can include some permeable layers, it
usually acts as an impermeable barrier. Foraminiferal limestones (E1)
are present in the Lower Eocene rocks which discordantly lay on the
Cretaceous rudist limestones (K22). Foraminiferal limestones can be
often poorly or sometimes well permeable depending on the degree
of fracturing, while the Cretaceous limestones are usually fractured
and weathered which means they are permeable, and they can be con-
sidered as the main aquifer in the survey area.
Krk is one of the biggest Adriatic islands where are several locations
with significant water resources. To the south of the study area there are
wells which exploit fresh water from sandstone layers within flysch
layers, as an exception in that region of Dinarides. The flysch layers in
the survey area are mainly impermeable, and the main goal is karstified
and fractured Cretaceous limestone in the basement. On the basis of test
pumping in the exploited wells on the Adriatic karstified islands, Terzić
(2006) estimated average hydraulic conductivity in the range 10−5–
10−6 m/s for rock masses.
50 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63
Fig. 3. Geological map (Mamužić et al., 1969) and positions of the profiles at the Baška area (island Krk) placed in the Dinarides.
The survey area is placed in the syncline structure (Fig. 3). Based on
previous investigations in the wider area we expect great variations in
the thicknesses of the flysch impermeable layers above the carbonate
basement as steeply sloping carbonate-clastic interface are suggested.
Thus, the predicted depth of the limestone aquifer ranges from 100 to
300 m.
3. Data acquisition
In the area of Gotalovec, investigations were carried out at two micro-
locations. The first one is located in a hilly area, and the expected depth of
Triassic dolomites is relatively small (tens to hundreds of meters) since in
the surroundings can be found outcrops of Triassic dolomites. The second
micro-location is in the plain area and expected depth of Triassic dolo-
mites is few hundred meters because the borehole in the vicinity of re-
search area revealed dolomites at 266 m (Pickel, 2004).
At the first micro-location ERT and CSAMT measurements were car-
ried out on the same profile. Electrical tomography profile was 700 m
long. The measurement was performed using a Wenner electrode
array with a unit electrode spacing of 10 m, so the depth of investigation
was 135 m. The inversion was done using the software based on the
Loke and Barker (1995) algorithm. Since expected depths are rather
small, and resolution of ERT to those depths is good, it can be used for
control and evaluation of CSAMT measurements. CSAMT measurements
were carried out at 16 CSAMT sounding sites distributed along the elec-
trical tomography profile. The distance between first three sites is 25 m,
and the others are at a distance of 50 m, so the total length of the profile
is 700 m.
At the second micro-location, deep vertical electrical soundings
using Schlumberger array (electrode spacing AB/2 = 1500 m) were
performed, covering a wider area in order to determine the general lith-
ological relationships, that is the common depth of Triassic dolomites.
Since the resolution of electrical methods is strongly reduced with
depth, seismic reflection measurements were performed. It enables
very high resolution even at depths of several hundred meters. The
aim of this investigation was to define more precisely discontinuities,
lithologic boundaries and potential faults. In particular, it was expected
to define more precisely the contact of clastic rocks and carbonates,
which is the relief of Triassic dolomites underlying clastic rocks. There-
fore, two reflection profiles, RF-1 and RF-2 were set up. Seismic reflec-
tion measurements were carried out using 24-channel ABEM
Terralock MK6 and standard CDP-method was applied. The data were
recorded using the in-line spread (source-receiver geometry) with an
offset of 20 m and 5 m geophone spacing. This resulted in a 12-fold cov-
erage with a CDP distance of 2.5 m. The source was small explosive
charge (100–200 g of explosive) placed in a shallow borehole about 1
m deep. The CSAMT measurements were carried out along seismic re-
flection profile RF-1, on 15 CSAMT stations at a distance of 25 m. The
electrical sounding results and reflection data allow detailed control
and analysis of CSAMT results.
In the area of Baška, measurements were performed at several
micro-locations and one has been selected as characteristic one
(Figs. 1 and 3). CSAMT measurements were completed on the profile,
and 13 CSAMT stations were measured with spacing about 50 m. ERT
measurements were done on the profile that stretches parallel to the
CSAMT-profile, using a Wenner electrode array with unit spacing of
 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63
10 m. Profile length is 800 m, so investigation depth is to 135 m. The in-
version was done using the method published by Loke and Barker
(1995).
The CSAMT method is very sensitive to noise caused by urban infra-
structure or surface inhomogeneities. Power lines, transformer stations
and other urban objects generate very high level of noise, so it is re-
quired to move away from them to an appropriate distance. Micro-
locations in the area of Gotalovec are located near a rural village, and
close to power lines, so measured data at some frequencies are strongly
corrupted. Therefore, prior to interpretation, the data have to be
smoothed, and some CSAMT stations even had to be discarded. This
step is essential since the retention of noisy data can very strongly influ-
ence on the results obtained by modelling, especially inversion. In the
area of Baška, the upper part of the profile is far from all the infrastruc-
tural facilities, so the urban noise is rather weak. Still, the noise in that
part of the profile is relatively strong. The noise can be explained by
karst terrain conditions because near the surface is situated a block of
carbonate rocks, characterized by large surface inhomogeneities. In-
creased level of noise is probably caused by groundwater flow through
the carbonate rocks that causes large variations of natural potential.
4. Natural and CSAMT data
The CSAMT system Geometrics STRATAGEM EH4 used in field mea-
surements allows frequencies up to 10 kHz to be recorded. The investi-
gation depth is frequency dependent and it is inversely related to the
square root of the frequency. According to the applied frequency
range and depending on the subsurface resistivity, the system can
reach depths between 500 m and 1000 m. This system uses also a con-
trolled source antenna to supplement natural signal in the range of 1
kHz to 70 kHz, thereby obtaining more reliable data and a better resolu-
tion for shallow features. This is due to the fact that natural field is espe-
cially strongly deformed due to the effect of various high-frequency
noise. The transmitter and the antenna for generating the artificial
field have to be placed at a particular distance from the profile in
order to approximate the signal with a plane wave electromagnetic
field. For this reason, the transmitter and the antenna are placed at dis-
tances greater than 400 m from the profile.
In a horizontally layered earth, where each layer is homogeneous
and isotropic, an electromagnetic wave propagates so that the electric
and magnetic field vectors are orthogonal (Cagniard, 1953). The ratio
of electric to magnetic field intensity (Ex/Hy) is a characteristic measure
of the electromagnetic properties called the characteristic impedance
(Z). The surface impedance is complex, frequency dependent and is a
tensor:




Ex Zxx Zxy Hx
¼ : ð1Þ
Ey Zyx Zyy HY
For the case of 2D conductivity structure, when measured in the co-
ordinate frame parallel to electric strike direction, the diagonal elements
of the impedance tensor Zxx and Zyy are zero (Bahr, 1988). The strike di-
rection is the horizontal direction along which the conductivity does not
change for 2D structure. Therefore the surface impedance is calculated
according to:
Ei
Zij ¼ ; ð2Þ
Hj
where i and j are in x and y direction. The apparent resistivity and im-
pedance phase are calculated from impedance components as:
 2 1
ρij ¼ Z ij  ð3Þ
ωμ 0
51
and
 !
−1 Im Z ij
ϕij ¼ tan   : ð4Þ
Re Z ij
For the homogenous half-space, the phase difference is equal to π/4.
Therefore, for the uniform resistivity the phase would be constant at all
frequencies and the indication of more complex subsurface structure is
the variation of phase and apparent resistivity with frequency.
Using the four channels Geometrics Stratagem EH4, electrical field
components (Ex, Ey) were measured on two perpendicular dipoles.
The dipole lengths were between 10 and 20 m, in order to acquire
good data quality. The horizontal magnetic field components (Hx, Hy)
were measured using induction coils (EMI BF-6). Recorded electric
and magnetic field components in two orthogonal directions are used
to calculate the surface impedance (Z) at a measurement site. The sur-
face impedance components are calculated using tensor processing
techniques and the apparent resistivity and phase curves are calculated
from impedance components.
In the case of 2D structures, the measured data can be separated into
two distinct modes, and these are referred to as E and H polarizations.
For E-polarization mode (TE) the electric field is polarized parallel to
the strike direction, and for H-polarization mode (TM) the magnetic
field is polarized in the strike direction. According to Berdichevsky
et al. (1998) both components TE and TM should be used in interpreta-
tion since they are complementary. They concluded that the TM mode is
more sensitive to near-surface structures and the TE mode usually to
deep structures. Also, the TM mode is more sensitive to the static shift,
while the TE mode may be almost undistorted. Another approach of
using both modes to obtain reliable solution is determinant (D) of the
impedance tensor as it is rotationally invariant for 2D structures.
Pedersen and Engels (2005) demonstrated that determinant impedance
is practical for use in 2D modelling since regardless of assumed strike di-
rection the data to be modelled are the same.
The measured apparent resistivity and phase data are presented in
Fig. 4 with two characteristic examples for each location. The surveys
are oriented so that the x-direction is parallel to the profile and y-
direction is perpendicular to it. Measured data already indicate the
structure as expected from previous investigations, electrical and seis-
mic explorations. At the first location in the Gotalovec area (MP-1) the
data are the noisiest (Fig. 4a). The apparent resistivity curves for yx di-
rection have different shape at frequencies lower than 3000 Hz which
could be the indication of dimensionality of the structure (2D or 3D).
The xy curve at station MT-6 shows increase of resistivity for lower fre-
quency that agrees with the model obtained by ERT (Figs. 4a and 5). At
the MT-10, located in the second part of the profile, the resistivity in-
creases at lower frequency as expected since high resistivities are not
obtained by measurement in this part of ERT profile. At the second loca-
tion (MP-2) the difference is not so great, while at the third location
(MP-3) resistivity curves for xy and yx direction have almost the same
values (Fig. 4b and c). At Gotalovec MP-2 on the station MT-5 phase
values are close to 45° for high frequencies to 5000 Hz, and also at
MP-3 to 4000 Hz for both stations indicating that the resistivity is al-
most uniform. After that the phase values decrease indicating an in-
crease of resistivity with depth.
5. Inversion and forward modelling of CSAMT data
The CSAMT data recorded along profiles were interpreted to obtain
resistivity models that reflect subsurface geology in two ways, using in-
verse and forward modelling. Both, the inverse and forward modelling
were performed using software for two-dimensional interpretation of
MT data ZondMT2D (Kaminsky, 2016).
An automatic inversion of observed data is faster interpretation
method to recover resistivity distribution in the underground. The
52 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63

Fig. 4. Examples of measured data: apparent resistivity and phase in the xy and yx direction plotted as a function of frequency, represented with two characteristic stations at each location:
(a) for the first location at Gotalovec, (b) for the second location at Gotalovec and (c) Baška on island Krk.

inversion is usually carried out by an iterative procedure where forward
modelling is included to calculate resistivity responses for initial
models. For the first initial model a half-space of constant resistivity is
assumed. The only influence of the interpreter is to define inversion al-
gorithm and several inversion parameters, such as resistivity limits,
smoothing factor and smoothness ratio. Depending on the data quality
and noises the inversion model can sometimes give unreliable results.
Also, we have to keep in mind that inversion is non-unique and many
resistivity models can fit the observed data. However, the inversion
model would provide for general structure of the subsurface based on
inverted resistivity distribution.
A forward modelling is done for the lithologically constrained model.
The model is defined using a priori geological data or results from other
geophysical methods. In that way the number of possible solutions is
significantly reduced as the interpreted model is under control of the
interpreter. During forward modelling it is possible to test several op-
tional models because sometimes different boundary depths and resis-
tivity contrasts are possible. Due to the fact that the final model needs to
fit both, a priori geologic knowledge and observed CSAMT data, the final
interpreted model will be often more reliable than the model acquired
by an automatic inversion.
The inversion could be done using just apparent resistivity data, just
phase or both. In the present analysis we have employed an Occam's in-
version (deGroot-Hedlin and Constable, 1990), with use of smoothing
operator and additional contrast minimization to obtain simple and
rather smooth resistivity model (Kaminsky, 2016). The input data
were apparent resistivity and phase data from Stratagem EH-4. Prior
to inversion, the CSAMT data need to be smoothed due to noise, since
noisy data may cause unrealistic features to appear in the model. The
data were inspected visually for each station in order to remove spikes
 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63 53

Fig. 5. 2D inversion of determinant data along profile MP-1 at the Gotalovec area (Pannonian basin). Comparison of observed and calculated apparent resistivity is presented at the top, and
ERT profile (TP-1) is in the middle. Top of high resistivity carbonate rocks at the western part of profile can be interpreted at the same depth in both models and the lateral contact is almost
vertical.

and smooth the data. The strike angle has been determined according to
geological trends and the data were rotated to TE (electric polarization)
and TM (magnetic polarization) mode. The measured data are rather
noisy and Jones and Groom (1993) showed that the strike angle deter-
mination based on field data is unstable in the presence of noise and
local distortion. Since the determination of strike angle from the data
themselves was questionable we have choose the aid of geological
trend.
The starting model for inversion was constructed with constant re-
sistivity and represented by a grid defined by the thickness of surface
layer, incremental factor and number of layers which determines the
depth of the bottom layer. The block width is scaled to match the site
spacing with two additional nodes between stations, and since the sta-
tions were evenly spaced it produced almost equidistant mesh. The
starting thickness of the first layer was set to 1 m and number of the
layers 38 to 40, depending on the profile. The incremental factor was
1.2, which means that each subsequent layer's thickness is multiplied
by 1.2 relative to the thickness of the layer above it. It resulted in the
depth of the bottom layer being between 5000 and 7000 m.
The computation was done by joint inversion of TE and TM data, and
also the determinant (D) which can be considered as the arithmetic
mean of TE and TM data. According to Pedersen and Engels (2005) the
determinant of the impedance tensor is independent of the assumed
strike direction so the best fitting model is found to be practically inde-
pendent of the starting model used. The inversion is done iteratively to
fit the calculated model to measured CSAMT data, until the given criteria
is reached, which can be either the maximum number of iterations or
the maximum RMS error. We have used a number of iterations as stop
criteria; in most cases 5 to 7, since a great number of iterations some-
times makes the model unnecessarily complicated and can result in a
model with unrealistic details. Our goal was to invert for the model
that will give a picture of the major structures.
Based on the inverted resistivity models and other available data, the
initial models for the forward modelling were constructed. The
54 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63

resistivity models were created in a block mode with the sites spacing
from the field layout and taking into account the topography. In forward
modelling it is approximated that the field in each cell of the mesh
changes linearly. The response (resistivity and phase) for defined resis-
tivity model is calculated for both modes and also determinant (D). Be-
cause shallow layers of low resistivity can significantly influence the
calculated data, the results from electrical resistivity tomography and
electrical soundings where available were used to define shallow layers.
In that way we could more reliably determine the deeper boundaries,
i.e. the interface between clastics and carbonate rocks.
6. Geophysical interpretation
6.1. Gotalovec area
At the first micro-location in the Gotalovec area, the Occam inversion
method was applied to obtain 2D resistivity model. Along the profile
MP-1 that is 700 m long, 16 CSAMT soundings were measured. The final
resistivity model was obtained after six iterations (Fig. 5). In the resulting
model, at the beginning of the profile resistivity is high (about 500 Ωm)
at a very shallow depth (20–30 m). At a distance between 200 and 250
m resistivity is sharply decreased, and the border is steep, which can al-
ready be seen on the ERT profile (Fig. 5). In the second part of the profile
(from 300 m to the end) resistivity is around 120 Ωm and contours
show a wide trough. But, comparing the observed and calculated apparent
resistivity sections the greatest discrepancy can be seen exactly for the low
frequency in the second part of the profile where the measured apparent
resistivities are higher (Fig. 5). To the depth of about 120 m CSAMT
model can be compared with the tomography profile TP-1 and agreement
is good. In the tomography model high resistivity can be observed at the
beginning of the profile at shallow depth (20 m), while in the second
part of the profile resistivities are low. But in the second part of the tomo-
graphic profile, the basement probably couldn't be reached because of
greater burial depth. Near surface the resistivity is very low (about 20
Ωm). High resistivities originate from carbonate rocks while low resistivi-
ties originate from impermeable clastic rocks. The lateral contact of high
and low resistivities is sharp and sub-vertical so a fault contact between
carbonates and clastics can be assumed.
Based on MT inversion results and ERT model, MT forward model-
ling was carried out. The ERT profile helped in defining the depth and

Fig. 6. Results of forward modelling along profile MP-1. At the top are observed and calculated apparent resistivity sections for the final model presented at the bottom. The unbroken line
represents the interface between clastics and carbonates in the final model. Two other possible solutions are presented with a dotted line and a dashed line.
 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63 55

lateral extent of shallow low-resistivity zone, which can significantly
influence the depth and shape of buried carbonate interface. The
CSAMT inversion has been used to define deep parts and carbonate bed-
rock in the initial resistivity model. A final solution of the forward
modelling has been selected according to the RMS error and previous
investigations. The final model shows very shallow carbonate rocks
(about 20 m) in the first 200 m of the profile (Fig. 6). Sharp, almost ver-
tical contact between high resistivity carbonates (800 Ωm) and clastics
is defined at this distance. The depth of the basement is much greater in
the second part of the profile and varies slowly, from 180 to 140 m, at a
distance 320–700 m. However, greater depths of about 220 m are defined
in the narrow zone at a distance 200–320 m. The resistivity of carbonates
changes laterally, and in the second part of the profile is 500 Ωm. If the car-
bonate interface is about 25 m shallower in the middle part of the profile,
than the deepest part should be about 50 m wider (Fig. 6). But, for the
narrower zone the depth could be as great as 280 m with small change
in RMS error.
Comparison of MT inversion and forward models gives similar re-
sults and very good agreement in the western part of the profile. But
the trough in the middle of the profile is more pronounced in the in-
verse model than in the forward model, and the transition from low to
high resistivity is quite wide so it is difficult to define the contact bound-
ary. The vertical uplift of the high-resistivity block at the beginning of
the profile also suggests a fault contact.
At the second micro-location in the area of Gotalovec electrical
soundings were measured in a wider area prior to CSAMT measure-
ments (Fig. 2). All electrical sounding curves show an increase in resis-
tivity in their deepest parts. The increase is caused by possible aquifers,

Fig. 7. 2D inversion of CSAMT data along profile MP-2.

56 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63

but the depth of the high resistivity body varies in a very wide range,
from 190 to 700 m (Fig. 2). An exploration borehole has been drilled
at the electrical sounding site with the shallowest interpreted depth
(GS-3). Exploration well was positive and confirmed the results of elec-
trical sounding. The Miocene limestones at a depth of 266.5 m, and
Triassic dolomites below them were determined in the borehole. The
water quality and recovery possibilities were better than expected. In
order to get more precise data about the burial depth of carbonate
rocks, the data from exploratory borehole have been used to calibrate
the resistivity, thus reducing the ambiguity of interpretation. At the

Fig. 8. Forward CSAMT model along profile MP-2 (Gotalovec area). At the top are observed and calculated apparent resistivity sections for the final model presented at the bottom.
Unbroken black line represents the step boundary between clastic and carbonate rocks in the final model. The shift in boundary depth of ±10% resulted in RMS error increase by
around 0.2% (dashed lines). The model with greater depth corresponds better with the depth from DC soundings (black dots).
 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63
sounding site GS-3 the depth of carbonate rocks has been determined at
266.5 m as a reliable interface so the resistivity of carbonates of 367 Ωm
was determined. According to these results, the resistivity of carbonates
was interpreted as 400 Ωm. Reinterpretation of other sounding curves
has been performed using this resistivity value, and more precise infor-
mation on the burial depth of carbonate rocks was obtained. In that way,
the depths of aquifers, Triassic dolomites and Lithothamnium lime-
stones, on a wider area were generally defined by electrical soundings.
The CSAMT data were measured at 15 stations along reflection pro-
file to acquire the information about resistivities. Water-bearing car-
bonates are clearly visible on the CSAMT-profile because of high
resistivities compared to impermeable deposits characterized by very
low resistivities (Figs. 7 and 8). The inversion of the CSAMT profile
MP-2 was performed using the Occam method and the final model on
Fig. 7 has been obtained after nine iterations (RMS = 9.5%). The depth
of carbonate rocks is difficult to define, but it could be interpreted as be-
tween the depths of 150 and 200 m. The resulting resistivity model does
not fit entirely the expected results, since the interpreted depth on the
nearby electrical sounding sites is much deeper (327 at GS-10 and
353 m at GS-9). Also, it is expected that the depth of interface becomes
shallower towards the end of the profile but the depth is the greatest
around a distance of 250 m. If the depth of interface is defined between
100 and 150 m than its trend in the first part of the profile is as expected.
The forward modelling shows much better agreement with other
available data (Fig. 8). An initial model was constructed based on inter-
pretation of DC soundings at GS-9 and GS-10. GS-9 is located 50 m from
MT-2 station and the profile, and GS-10 is located on the CSAMT-profile
at a distance of 215 m (Fig. 1). The top of carbonate rocks has been
interpreted at the depth of 353 m at GS-9, while at GS-10 it is at the
depth of 327 m. In the initial 2D model the depth of carbonates was
350 m and has been constant to the distance of 150 m. After that the
Fig. 9. CSAMT and seismic reflection profiles at the Gotalovec survey area. Contact of clastic and carbonate rocks is defined more precise on the basis of seismic reflection and electrical
sounding data.
57
depth gradually decreased till the end of the profile, where it was 225 m.
The resistivity of carbonates was at first defined to 400 Ωm, as determined
by electrical soundings. In the next step the resistivity has been increased
to 650 Ωm in order to reduce the RMS error (9.7%).
The calculated apparent resistivity section for model with clastic resis-
tivity of 55 Ωm and carbonate rock resistivity of 650 Ωm compared to ob-
served apparent resistivity section shows good agreement (RMS = 4.5%).
However, if the top of the carbonates is 25 m shallower compared to the
starting model, the RMS error is 8.1%. Thus, the final model shows the car-
bonate rocks depth of 325 m at the beginning of the profile and 200 m at
the end of the profile, presented with a full line over the smoothed final
model in the background (Fig. 8).
Since the resolution strongly decreases with depth, uncertainty anal-
ysis has been performed within the forward modelling. If the depth of
carbonate boundary is shifted for −10%, which means it is shallower
for 20–33 m in regard to the final model, the fit to the observed data
is almost the same with RMS error 4.5%. Similar results were obtained
when the boundary depth was greater for 10% (RMS = 4.8%), Fig. 8. If
the depth of carbonates is even greater, i.e. 400 m at the beginning of
the profile and 275 m at the end of the profile, the RMS error is now in-
creased (5.2%). The fit is also less in the case of the model with a larger
jump i.e. great change in the depth of carbonate boundary at a distance
of 150 m (Fig. 8; dotted line). Therefore, based on the forward model-
ling it can be assumed that the depth of carbonate rock boundary is re-
solved to a depth of ±10%.
Additionally, more precise reflection measurements were per-
formed at the identified location, which made it possible to precisely de-
fine lithological relationships. On the reflection profile, three sets of
strong, mostly continuous reflections were noted in terms of depth, be-
tween which there are zones with weak and discontinuing reflections
(Fig. 9). At the beginning of the profile, there is the first set with a two-
58 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63

way travel time of 140 ms, the second one is in the 180–300 ms interval,
and the third one in the 355–455 ms interval. If the boundary between
marl and carbonates is introduced from the GS-9 electrical sounding
(Fig. 9), and assuming the average seismic velocity of 2500 ms, the inter-
face will be set to the deepest reflection in the second set which is located
at two-way time of 310 ms at the beginning of the profile. By following
this reflection towards the end of the profile, the time of 180 ms can be
read. Since it is a well-known fact that clear reflections cannot be obtained
in carbonates, the interface between marl and carbonates defined in this
manner is completely logical. When transferring this horizon to the trans-
versal reflection profile P-2, it can be observed that the same boundary is
again the last deepest reflection of the second set, under which extends a
zone of weak and discontinuous reflections.
Results of the MT forward modelling show very good agreement
with the seismic reflection data (Fig. 9). At the beginning of the profile
the depth is the greatest (325 m), and varies very little to the middle
of the profile, but then rapidly decreases towards the end of the profile
(Fig. 8). Mapping of carbonate interface can be more precisely carried
out in the reflection data (Fig. 9). Because of the lower resolution of
the CSAMT method, the boundary cannot be as precisely defined as on
the reflection profile. However, in terms of hydrogeology, the CSAMT
method generally provides more complete data. Based on resistivities,
conclusions can be drawn about the existence of an aquifer and its burial
depth. On the other hand, seismic reflection provides more precise data
about discontinuities and relative structural relationships, but it does
not say anything about rocks at the contact of discontinuities, so its

Fig. 10. Resulting model of 2D inversion for profile MP-3 at the Baška (Dinarides); bottom. At the top are pseudo sections of measured and calculated apparent resistivity and in the middle
is ERT model along profile measured parallel to MP-3. Near surface high-resistivity body is clearly visible on both models, and to a depth of about 100 m resistivity is low.
 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63 59

potential can be completely explored only when it is combined with
electrical sounding and audio-magnetotellurics. Therefore, in such a
model, the CSAMT method should be applied first in order to define po-
tential aquifers and indicate the most favourable micro-locations where
seismic reflection can be applied as the latter allows for a more precise
mapping of aquifer depths and potential fault zones.
6.2. Baška area
In the third micro-location on island Krk, the Cretaceous limestones
are not affected on the tomographic profile, even in the central part with
the greatest depth of the penetration, which means the depths are
greater than 100 m (Fig. 10). But the surface block of high resistivities,
with average thickness of 20 m, is clearly visible in the ERT data. This
block is also introduced in the MT forward modelling, and contributed
to reduce the interpretation ambiguities to define the depth of the lime-
stone aquifer.
The inversion of CSAMT data was performed using the Occam method
and the final model was obtained after five iterations for combined TM
and TE mode (Fig. 10). The inversion of determinant (D) data gives very
similar results with 10–20 m greater depth in the central part of the pro-
file. High resistivity originates from a carbonate basement. The model
shows high resistivities at a depth of about 120 m (isoline of approx.
100 Ωm) to a distance of 200 m, and then the depth gradually increases
to about 150 m until a distance of 400 m. The boundary is shallower to-
wards the end of the profile, and at the end is approximately 70 m. A shal-
low high resistivity body is also observed to a depth of no more than 25 m,
in a distance range from 20 to 250 m.
The initial model for the forward modelling has been constructed
using the ERT results and the inversion CSAMT results. A shallow high re-
sistivity body at a distance from 0 to 280 m and to a depth of 20 m was

Fig. 11. CSAMT forward modelling in the Baška area (Dinarides). The observed and calculated apparent resistivity pseudo-sections are displayed at the top. The unbroken black line
represents contact of clastics and carbonates in the final model which is shown in the background as smooth model. The second possible model, with constant depth of boundary is
indicated with dashed line.
60 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63
introduced, as well as the high resistivity body in the background of the
clastics with resistivities about 20 Ωm. Very good agreement of calculated
apparent resistivities and observed data (RMS = 8.2%) is obtained for the
model with gradual increase of carbonates depth from 75 to 115 m, up to
the distance of 280 m. The greatest depth (around 120 m) is in the central
part of the profile, till a distance of 460 m. To the end of the profile, the
slope of carbonate interface increases, and the depth decreases to about
60 m (Fig. 11). However, almost as good agreement (RMS = 8.4%) is ob-
tained for a somewhat different model, with constant depth of carbonates
(110–120 m) from the distance of 280 m to the eastern end of the profile.
But at the end of the profile, a body of 200 Ωm needs to be defined above
the carbonate rocks (Fig. 11).
It can be shown that the interface depth in this model is determined
within an accuracy of ±10 m. When the carbonate boundary in the final
model is shifted 10 m shallower, comparison of calculated and mea-
sured resistivity gives an RMS error of 8.9%, and if the boundary is
shifted 10 m deeper, the RMS error is 8.3%. But, if the shift is larger
and amounts 20 m, the RMS error is significantly increased and is 9.8%
for +20 m and 8.5% for −20 m.
Both CSAMT models, inverse and forward, show the similar configura-
tions (Figs. 10 and 11). In the first half of the profile both models show shal-
low high resistivity block (110 Ωm), which originates from weathered
limestones within the flysch deposits. The layer of lowest resistivity
(around 20 Ωm) extends to the depth of about 100 m and generally follows
the relief of the profile, except at the end of the profile where its thickness is
reduced. The low resistivities originate from impermeable package of flysch
deposits dominated by clay and marl. Below them there are high resistivi-
ties (500 Ωm) arising from fractured, permeable carbonate rocks.
However, it must be emphasized that the forward model provides a
significantly distinctive image than the inverse model. It is difficult to
define the boundaries in the inverse model since the area of transition
from low to high resistivity is very wide. Besides, a significantly lower
resistivity in the basement was observed at a distance of 300–450 m,
which could indicate a fault zone. But in the context of forward
Fig. 12. Geological cross-section along the profile MP-1 at Gotalovec.
modelling there is no need for introducing such a zone, so the fault
zone in the basement was not assumed.
7. Geological models
In the area of Gotalovec, low resistivities near the surface and ex-
tending to the greater depth in some parts of the profile, are interpreted
as the effect of impermeable clastic rocks. High resistivity below them is
caused by fractured and weathered carbonates which act as an aquifer.
Both resistivity models, the inverse and the forward, along profile MP-1
(Figs. 5 and 6) suggest a fault contact at a distance of 200 m. The contact
is subvertical so a normal fault with a great jump (about 200 m) is as-
sumed (Fig. 12). The lowest resistivity (about 20 Ωm) is near surface
and these are interpreted as clay and marl deposits. Between those
layers and Triassic dolomites in the basement there is an alteration of
impermeable and poorly permeable Neogene clastic layers: marl,
sandy marl and clayey sandstones. This model shows that great jumps
of faults can be expected in the area as a consequence of tectonic activ-
ities. It will help in the consideration of hydrogeological potential of the
selected micro-location.
At the profile P-2 the depth of carbonate interface is greater, espe-
cially at the beginning of the profile where it is greater than 320 m
(Fig. 13). Resistivity models and seismic reflection profiles clearly
show uneven carbonate bedrock, and the depth decreases from the be-
ginning of the profile towards the end (Figs. 7–9). But, high resistivities
here may originate from Miocene Lithothamnium limestones and Trias-
sic dolomites, which cannot be resolved based on the interpreted resis-
tivity. As both of these rocks represent aquifers, it is essential to define
the boundary of clastics and carbonates. So, although on the profile
are defined Triassic dolomites, in some parts of the profile also can be lo-
cated Miocene limestone in reef development. The faults are interpreted
on the basis of clear lateral breaking of reflections on the seismic profile.
The geological model shows great depth variations of clastic–carbonate
interface, that is the aquifer depths. This makes it possible to determine the
 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63
Fig. 13. Geological cross-section along profile MP-2 at the second micro-location in the
Gotalovec area.
Fig. 14. Geological cross-section along the MP-3 profile at the Baška area on island Krk.
61
well position at the shallowest aquifer, which will reduce exploration
costs. On the other side, the model also shows fault zones in the aquifer
which point to higher permeability, due to increase of secondary porosity,
where larger amounts of water can be drawn.
Near surface clastic rocks are mainly clays because the resistivity is
very low, while below them can be assumed four packages of clastic
layers based on sets of reflections on the seismic profile. The shallowest
set consists of alterations of marl, shale and clayey sandstone because
reflections are very strong, while the second set is composed of monot-
onous series of marl as reflections are weak (Fig. 13). The third set with
strong reflections is probably similar to the first set, so there is assumed
alteration of layers with different lithology: marl, shale and clayey sand-
stone. The fourth set, near the carbonate bedrock, shows relatively
strong but discontinuous reflections, particularly at the beginning of
the profile where the carbonates are the deepest. Therefore, also the lat-
eral changes in lithology can be assumed, and along the aforementioned
marl and shale, the presence of conglomerates, clayey limestones and
calcitic marl is possible.
The depth of carbonate basement on the profile P-3 in the area of
Baška generally follows the relief of the terrain and is about 100 m, ex-
cept at the end of the profile where the depth is significantly reduced
(Fig. 14). This reduction of depth is interpreted as a folding of layers in
the basement, that is, as a local anticline, although it might be caused
by tectonics. On the carbonate basement there is probably a relatively
thin package of foraminiferal limestones that are discordantly deposited
on Cretaceous limestones. Unfortunately, based on the resistivity model
Eocene limestones cannot be resolved from the Cretaceous limestones
(Fig. 11). Consequently, there is no steep slope of carbonate interface,
as is predicted based on the previous research in the wider study area.
A package of clastic flysch deposits has a low resistivity, so it is domi-
nated by marl, clayey marl and clayey sandstone. In the first half of
the profile there is a plate of carbonate rocks near surface where high re-
sistivities are interpreted, which is even more clearly recognized on the
tomography profile (Fig. 10). These limestones are very clearly visible
on the outcrops around the profile, and belong to the Middle Eocene
flysch package.
Based on the surface geological mapping and previous data in the sur-
roundings someone would expect relatively steep inclination of the
62 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63
flysch carbonate interface. In that case an exploratory borehole should be
placed as close as possible to this boundary at the surface, which is high
in the hill, to catch the carbonates at acceptable depth. But, the models
based on the CSAMT data shows that the interface shape is quite different
and the shallowest carbonate depth can be expected in the valley which
means the costs of the exploratory borehole will be reduced.
8. Conclusions
The CSAMT method provides the most complete data on lithological
and structural relationships in cases of hydrogeological targets that are
deeper than 100 m. It can reach targets at larger depths with dense
space sampling at significantly lower costs. In the mapping of carbonate
rocks underneath thick clastic impermeable deposits, the CSAMT
method provides information on the existence of aquifers and their gen-
eral burial depths, which can then be more precisely determined by
using seismic reflection on a selected potential micro-location. The in-
terpretation ambiguity of the CSAMT inversion methods, especially at
greater depths, can be significantly reduced by means of near surface
electrical resistivity tomography data. Namely, near surface thin layers
of low or high resistivities, which can be precisely determined by the
electrical resistivity tomography, may strongly affect the interpretation
depths of the target rocks.
On the selected micro-locations, Gotalovec in the area of Pannonian
basin and Baška in the Dinaridic area that are representing common
hydrogeological model in both regions of Croatia, CSAMT data together
with data of other geophysical methods (electrical resistivity tomogra-
phy, electrical sounding and seismic reflection) enabled the definition
of a reliable prognostic geological model. Great variations of the dolo-
mitic aquifer depths were determined in the Gotalovec area as a conse-
quence of a strong tectonic activity in the geological past. In the Baška
area depth changes of a limestone aquifer are mainly caused by the
layer folding.
Based on the presented models we can conclude that CSAMT
method can provide much greater depth than electrical resistivity to-
mography and can be considered as a fundamental geophysical method
for hydrogeological investigations, deeper than 100 m. It can provide
dense space sampling at much lower costs of research, so it has a
great advantage over classical electrical sounding, which, likewise,
may provide great penetration depth. For comparison it can be stated
that 8–12 CSAMT-soundings can be measured daily, while at the same
time only 1 or 2 deep electrical soundings can be measured. But, a
CSAMT research may demonstrate its advantages only in the case of
very dense layout of CSAMT stations, due to the greater sensitivity to
noise in relation to resistivity methods. Depending on the depth of in-
vestigation and the geological model, the distance between CSAMT sta-
tions along profile may be 25 to 100 m.
Interpretation of CSAMT data is much more complex in relation to
resistivity methods. Automatic inversion of 2D or 3D electrical resistiv-
ity tomography data (Loke and Barker, 1995, 1996) is a very powerful
tool that often provides reliable inverse resistivity models that can be
rather easily geologically interpreted. However, automatic inversion of
CSAMT data cannot sometimes provide for an accurate underground re-
sistivity distribution. In the considered examples the depth of carbonate
aquifers in the models obtained by inversion is generally greater than
the actual depth. More reliable models are obtained by forward model-
ling which enables including a priori data acquired by other geophysical
methods. Seismic reflection with significantly higher resolution than
CSAMT method can give great contribution to research results. In this
combination the seismic reflection provides accurate mapping of dis-
continuities (faults and layer boundaries), while CSAMT method can de-
fine lithology and estimate porosity based on interpreted resistivity, but
both methods assure great depth of investigation. Electrical resistivity
tomography can help in the accurate mapping of shallow resistivity var-
iations, which are difficult to detect using the CSAMT data because of
sensitivity to noise in the high frequency range.
Acknowledgements
We would like to thank Saša Kolar for his extraordinary contribution
to data acquisition and processing. We also thank Darko Dudjak and
Nataša Balaško for their contribution to some CSAMT measurements.
References
Bahr, K., 1988. Interpretation of the magnetotelluric impedance tensor: regional induction
and local telluric distortion. J. Geophys. 62, 119–127.
Berdichevsky, M.N., Dmitriev, V.I., Pozdnjakova, E.E., 1998. On two-dimensional interpre-
tation of magnetotelluric soundings. Geophys. J. Int. 133:585–606. https://doi.org/
10.1046/j.1365-246X.1998.01333.x.
Borović, S., Marković, T., Larva, O., Brkić, Ž., Mraz, V., 2016. Mineral and thermal waters in
the Croatian part of the Pannonian basin. In: Papić, P. (Ed.), Mineral and Thermal Wa-
ters of Southeastern Europe, Environmental Earth Sciences https://doi.org/10.1007/
978-3-319-25379-4_2.
Cagniard, L., 1953. Basic theory of the magnetotelluric method of geophysical prospecting.
Geophysics 18, 605–635.
Dahlin, T., 1996. 2D resistivity surveying for environmental and engineering applications.
First Break 14 (7), 275–283.
Dragičević, I., Blašković, I., Mayer, D., Tomljenović, B., 1998. Mountain aquifers of
the northern Croatia. XIXth Conference of the Danubian Countries on Hydrolog-
ical Forecasting and Hydrological Bases of Water Management, Osijek, Paper
No. 3.28.
Engelsfeld, T., Šumanovac, F., Krstić, V., 2011. Classification of near-surface anomalies in
the seismic refraction method according to the shape of the time–distance graph: a
theoretical approach. J. Appl. Geophys. 74:59–68. https://doi.org/10.1016/j.
jappgeo.2011.03.004.
Engelsfeld, T., Šumanovac, F., Pavin, N., 2008. Investigation of underground cavities in a
two-layer model using the refraction seismic method. Near Surf. Geophys. 6:
221–231. https://doi.org/10.3997/1873-0604.2008017.
deGroot-Hedlin, C., Constable, S., 1990. Occam's inversion to generate smooth, two-
dimensional models from magnetotelluric data. Geophysics 55, 1613–1624.
Ismail, N., Schwarz, G., Pedersen, B.L., 2015. Investigation of groundwater resources using
controlled-source radio magnetotellurics (CSRMT) in glacial deposits in Heby, Sweden.
J. Appl. Geophys. 73:74–83. https://doi.org/10.1016/S0926-9851(01)00044-1.
Jones, A.G., Groom, R.W., 1993. Strike-angle determination from the magnetotelluric im-
pedance tensor in the presence of noise and local distortion: rotate at your peril!
Geophys. J. Int. 113, 524–534.
Kaminsky, A., 2016. Program for two-dimensional interpretation of data obtained by
magnetotelluric sounding – ZONDMT2D. Zond Geophysical Software.
Loke, M.H., Barker, R.D., 1995. Least-squares deconvolution of apparent resistivity
pseudo-sections. Geophysics 1682–1690.
Loke, M.H., Barker, R.D., 1996. Rapid least-squares inversion of apparent resistivity
pseudo-sections by a quasi-Newton method. Geophys. Prospect. 44, 131–152.
Mamužić, P., Milan, A., Korolija, B., Borović, I., Majcen, Ž., Brkić, M., 1969. Osnovna
geološka karta SFRJ, list Rab, L33-114 (Basic geological map of SFRY, The Rab
sheet). Institut za geološka istraživanja Zagreb, 1959-1965. Savezni geološki zavod,
Beograd.
Pedersen, L.B., Engels, M., 2005. Routine 2D inversion of magnetotelluric data using the
determinant of the impedance tensor. Geophysics 70:G33–G41. https://doi.org/
10.1190/1.1897032.
Pedrera, A., Martos-Rosillo, S., Galindo-Zaldívar, J., Rodríguez-Rodríguez, M., Benavente, J.,
Martín-Rodríguez, J.F., Zúñiga-López, M.I., 2016. Unravelling aquifer-wetland interac-
tion using CSAMT and gravity methods: the Mollina-Camorra aquifer and the Fuente
de Piedra playa-lake, southern Spain. J. Appl. Geophys. 129:17–27. https://doi.org/
10.1016/j.jappgeo.2016.03.018.
Pickel, J., 2004. Dodatni vodoistražni radovi za potrebe zahvaćanja termalne vode na
području Gotalovca (Additional groundwater exploration in the purpose of thermal
water capture in the Gotalovec area) (Fond struč. dok. Geoida Beroš, d.o.o.).
Pous, J., Ledo, J., Marcuello, A., Daignières, M., 1995. Electrical resistivity model of the crust
and upper mantle from a magnetotelluric survey through the central Pyrenees.
Geophys. J. Int. 121:750–762. https://doi.org/10.1111/j.1365-246X.1995.tb06436.x.
Semenov, V.Y., Pek, J., Ádám, A., Jóźwiak, W., Ladanyvskyy, B., Logvinov, I.M., Pushkarev, P.,
Vozar, J., 2008. Electrical structure of the upper mantle beneath Central Europe: re-
sults of the CEMES project. Acta Geophys. 56:957–981. https://doi.org/10.2478/s
11600-008-0058-2.
Šimunić, A., Pikija, M., Hećimović, I., 1982. Osnovna geološka karta SFRJ, List Varaždin,
L33-69 (Basic geological map of SFRY, The Varaždin sheet). Geološki zavod – OOUR
za geologiju i paleontologiju Zagreb, 1971–1978. Savezni geološki zavod, Beograd.
Spichak, V., Manzella, A., 2009. Electromagnetic sounding of geothermal zones. J. Appl.
Geophys. 68 (4), 459–478.
Šumanovac, F., 2006. Mapping of thin sandy aquifers by using high resolution reflection
seismics and 2-D electrical tomography. J. Appl. Geophys. 58:144–157. https://doi.
org/10.1016/j.jappgeo.2005.06.005 (and Erratum to Mapping of thin sandy aquifers
by using high resolution reflection seismics and 2-D electrical tomography. J. Appl.
Geophys. 59 (4), 345–346).
Šumanovac, F., Urumović, K., Dragičević, I., 2003. Hydrogeological mapping of a Miocene
aquifer by two-dimensional electrical imaging. Rudarsko-geološko-naftni zbornik 15,
19–29.
Šumanovac, F., Weisser, M., 2001. Evaluation of resistivity and seismic methods for
hydrogeological mapping in karst terrains. J. Appl. Geophys. 47:13–28. https://doi.
org/10.1016/S0926-9851(01)00044-1.
 F. Šumanovac, J. Orešković / Journal of Applied Geophysics 153 (2018) 47–63
Terzić, J., 2006. Hidrogeologija jadranskih krških otoka (Hydrogeology of Adriatic
karstified islands). (Ph.D. Thesis). University of Zagreb.
Terzić, J., Šumanovac, F., Buljan, R., 2007. An assessment of hydrogeological parameters on
the karstic island of Dugi Otok, Croatia. J. Hydrol. 343:29–42. https://doi.org/10.1016/
j.jhydrol.2007.06.008.
Unsworth, M., Wenbo, W., Jones, A.G., Li, S., Bedrosian, P., Booker, J., Sheng, J., Ming, D.,
Handong, T., 2004. Crustal and upper mantle structure of northern Tibet imaged
with magnetotelluric data. J. Geophys. Res. 109:B02043. https://doi.org/10.1029/
2002JB002305.
63
Vlahović, T., Šumanovac, F., 2015. Defining a general hydrogeological model for Susak
Island, Adriatic Sea, Croatia. Q. J. Eng. Geol. Hydrogeol. 48:135–146. https://doi.org/
10.1144/qjegh2013-044.
Volpi, G., Manzella, A., Fiordelisi, A., 2003. Investigation of geothermal structures by
magnetotellurics (MT): an example from the Mt. Amiata area, Italy. Geothermics
32, 131–145.
Vozoff, K., 1972. The magnetotelluric method in the exploration of sedimentary basins.
Geophysics 37, 98–141.