Content uploaded by Yunhe Liu
Author content
All content in this area was uploaded by Yunhe Liu on Mar 31, 2025
Content may be subject to copyright.
Case History
High-resolution imaging of sedimentary bauxite deposits and structural
controls on mineralization in western Guangxi Province based on efficient
3D TEM observation and inversion
Laonao Wei1, Yunhe Liu1, Changchun Yin1, Bo Zhang1, Xiuyan Ren1, Yang Su1, and Zhihao
Rong1
ABSTRACT
Bauxite plays a crucial role in metallic and nonmetallic indus-
tries. The surface-exposed salento-type bauxite deposits have
been largely exploited and developed. With the increasing de-
mand for these resources, it is important but very challenging to
explore the potential bauxite deposits in the deep earth. In this
paper, based on new developments in transient electromagnetic
(TEM) technologies, we conduct explorations of the sedimen-
tary bauxite in Dengli-Tianyanggumei and Dajia mining areas in
western Guangxi Province. To achieve a fast and high-resolution
inversion, we adopt an array-based observation strategy for
large-scale 3D TEM and collect electromagnetic (EM) data
inside and outside the transmitting loop. Compared with tradi-
tional TEM surveys, the observation strategy can quickly ac-
quire the data for large-scale surveys and improve the data
acquisition efficiency by more than 25 times. We then use a 3D
inversion algorithm to estimate the underground conductivity
structure and analyze the distribution of the sedimentary baux-
ite. To do that, we discretize the undulating surface and trans-
mitter-receiver locations with unstructured grids and use the
finite-element and quasi-Newton methods to achieve high-res-
olution imaging of subsurface electrical structures. Since the
observation strategy greatly reduces the number of transmitters,
the efficiency of 3D EM inversions can be significantly
improved. Experiments over two mining areas show that our
inversions can clearly recover the underground resistivities.
The inferred burial depth and spatial distribution of the sedimen-
tary bauxite are in agreement with the drilling data. By combin-
ing electrical resistivity tomography results and geologic data,
we illustrate the impact of faults on the spatial distribution of
potential sedimentary bauxite deposits.
INTRODUCTION
Bauxite is a sedimentary rock that is the main source of alumi-
num and gallium. It has a wide range of applications in the cement
and steel industry, metallurgy, refractory materials, etc. (Cablik,
2007;Gawu et al., 2012). The origin of bauxite deposits mainly
includes lateritic (salento-type) and sedimentary types (Bogatyrev
and Zhukov, 2009;Liu et al., 2010;Mongelli et al., 2015). The sale-
nto-type bauxite deposits are mostly exposed at the earth surface
and easy to exploit, which dominates the supply of bauxite deposits
at present. However, with the increasing demand for bauxite, the
salento-type deposits are approaching depletion, and the concealed
sedimentary bauxite deposits in the underground have gradually be-
come the focus of exploration and extraction efforts. Over the past
three decades, various geophysical exploration methods have been
applied to explore bauxite deposits, such as shallow seismic reflec-
tion (Ayman et al., 2015), electrical resistivity tomography (ERT)
(Nogueira et al., 2011), controlled-source audio-frequency magne-
totelluric method (CSAMT) (Wei et al., 2011) or transient electro-
magnetic (TEM) methods (Zhang, 2005).
Manuscript received by the Editor 2 November 2023; revised manuscript received 13 December 2024; published ahead of production 6 January 2025;
published online 24 March 2025.
1Jilin University, College of Geo-Exploration Science and Technology, Changchun, China. E-mail: weiln21@mails.jlu.edu.cn; liuyunhe@jlu.edu.cn
(corresponding author); yinchangchun@jlu.edu.cn; zhangbo@jlu.edu.cn; renxiuyan@jlu.edu.cn; suyang@jlu.edu.cn; rongzh22@mails.jlu.edu.cn.
© 2025 Society of Exploration Geophysicists. All rights reserved.
B131
GEOPHYSICS, VOL. 90, NO. 3 (MAY-JUNE 2025); P. B131–B142, 14 FIGS., 3 TABLES.
10.1190/GEO2023-0649.1
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
The use of shallow seismic reflections is a mainstream method
for bauxite exploration. Dragičevićet al. (1991) use it in the
Mešihovina-Rakitno area (western Bosnia) to locate new bauxite
deposits. Nogueira et al. (2011) apply seismic refraction in the
Barro Alto region of Goiás, Brazil, to determine the interface be-
tween bauxite layers and syenite. Ayman et al. (2015) use high-res-
olution seismic reflection in the Khashm region of Saudi Arabia to
image bauxite layers. Xue et al. (2020) explore the concealed sedi-
mentary bauxite deposits in the Mianchi area of western Henan
Province, China. Shallow seismic reflections can usually achieve
high-resolution imaging and delineate boundaries between bauxite
layers and underlying bedrocks. However, since bauxite deposits
are usually very thin in Guangxi, it is difficult to effectively identify
the deep-seated ore bodies in most cases with this method. In recent
years, passive seismic methods have also been applied to bauxite
exploration (Polychronopoulou et al., 2020;Orfanos et al., 2021)
but this resolution is lower than with shallow reflection methods
and requires longer data acquisition time.
Because bauxite deposits are usually more conductive than the
surrounding rocks, the ERT and other direct current (DC) resistivity
methods are widely used in bauxite exploration. Bi (2009) applies
ERT to detect concealed bauxite deposits in the western Henan
Province. Nogueira et al. (2011) use the DC resistivity method
in the Barro Alto region of Goiás, Brazil, to delineate the interfaces
between syenite and bauxite based on resistivity distribution. Huang
et al. (2018) achieve the indirect detection of sedimentary bauxite
deposits in the western Guangxi using ERT. Xue et al. (2022a) use
the DC resistivity method in the Wanggudong area of western
Henan Province, China, and obtain the burial depth and horizontal
position of sedimentary bauxite deposits. Although the DC methods
have high accuracy in shallow detection, it has limitations in
detecting the deep-seated bauxite deposits.
The CSAMT method is also extensively used in bauxite explo-
ration. Wei et a l. (2 011 ) use the CSAMT method in the Shisi-
Beiyu bauxite area of Xin’an County and successfully determine
the interfaces of the paleo topography with Ordovician limestone.
Li et al. (2013) use it along with audio-frequency magnetotelluric
(AMT) methods to explore the Chepan mining area of Wulong,
inferred the spatial distribution of concealed bauxite deposits.
Yang et al. (2019) use the CSAMT method to determine the spatial
distribution of bauxite deposits in the Yushan-Lizhuangzhai area
of Henan Province. Hao (2022) infers the locations and depths of
bauxite mineralization based on the variations of the Ordovician
erosion surface detected by CSAMT. Although the CSAMT
method has a large exploration depth, its resolution is low com-
pared with ERT and TEM. This means that the CSAMT method
cannot be used to directly image the bauxite deposits and thus is
mostly used to explore the background structures. The passive
AMT method was also adopted to image bauxite deposits. Guo
and Lu (2015) achieve good results in the exploration of the
Tuanxi bauxite in Guangxi using the AMT method. Similarly,
Zhang et al. (2015) delineate mineralized sites of bauxite deposits
in the Jinchanggou area of Guizhou Province. However, the AMT
method has usually low signal-to-noise ratio, resulting in an even
lower resolution than the CSAMT method. The detection depth
of DC resistivity methods is relatively shallow and can only detect
anomalies to a depth of several tens of meters. The CSAMT can
detect up to a depth of hundreds of meters, whereas the depth of
AMT detection can reach thousands of meters.
The TEM method offers high resolution, relatively large explo-
ration depth at low cost (Xue et al., 2008). It has a clear advantage
over DC resistivity and ground electromagnetic (EM) methods such
as CSAMT or LOTEM/SOTEM for detecting horizontal conduc-
tors. This makes TEM particularly well suited for exploring new
sedimentary bauxite deposits (Maher, 1992;Guo et al., 2020).
Zhang (2005) demonstrates that TEM can provide information
on the topographic surface of the Ordovician limestone and the lo-
cation of surface depressions for the detection of deep-seated baux-
ite deposits. Zhang (2007) infers the burial depth and stratigraphic
interface variations of the interfaces between the bauxite-bearing
rocks and underlying limestone using resistivity profiles obtained
from TEM inversion. Wang et al. (2010) successfully distinguish
the surfaces of a highly resistive basement using the TEM method
and delineated favorable areas for bauxite mineralization. Li et al.
(2019) conduct TEM surveys over a bauxite deposit in Henan Prov-
ince and infer the approximate positions of ore-bearing layers and
mined-out areas. Lai et al. (2021) use the equivalent magnetic flux
TEM to detect bauxite deposits and achieved accurate lithostrati-
graphic divisions. Xue et al. (2022b) infer the locations and depths
of bauxite deposits based on resistivity differences obtained by
TEM method.
The TEM method is highly effective for bauxite exploration.
However, TEM data inversion of strategies are mainly one- or
two-dimensional with strong ambiguity and insufficient resolution,
making them unsuitable for effective exploration of sedimentary
bauxite deposits with complex topography and geologic settings.
In this study, based on high-resolution 3D inversion algorithms de-
veloped by Liu et al. (2019) and Zhang et al. (2021) for large trans-
mitter loop TEM (Nabighian and Macnae, 1991;Dentith and
Mudge, 2014), we conduct TEM surveys in two sedimentary baux-
ite mining areas in western Guangxi Province. Our goal is to iden-
tify effective methods for efficient detection and interpretation of
sedimentary bauxite deposits.
GEOLOGIC BACKGROUND
There are abundant bauxite deposits in counties such as Jingxi,
Debao, Tiandong, and Pingguo in Guangxin Province, China
(Zhang, 2011). The bauxite belt in western Guangxi Province
is mainly composed of Quaternary, Tertiary, Triassic, Permian,
Carboniferous, Devonian, and Cambrian formations (Figure 1).
The mining area mainly consists of two types of deposits: the sale-
nto-type and sedimentary bauxite deposits. The sedimentary baux-
ite deposits occur between the bottom of the Late Permian Heshan
Formation and the middle Permian Maokou limestone Formation,
above the ancient erosion surfaces. They are in parallel uncon-
formable contact with the underlying Maokou limestone Forma-
tion, forming distinct layered structures and serving as the primary
source of salento-type bauxite deposits. The salento-type bauxite
deposits are found in the Quaternary red earth deposits of Neogene
system (Wei , 199 9). The mining area exhibits developed fault
and fold structures, mainly trending northwest. The sedimentary
bauxite deposits are generally distributed in the northeast direction
in Jingxi-Duyangshan uplift zone of Youjiang fold belt (Qin
et al., 2019).
The survey areas considered in this work are located in the Debao
County and Jingxi City. The Dengli-Tianyanggumei mine in Debao
County is located at the border between the Debao and Tianyang
counties, as shown in Figure 1a. The ore body is layered with a
B132 Wei et al.
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
simple internal structure without any interlayers. The thickness of the
ore body is relatively stable, and it is generally continuous, with local
discontinuities due to faulting. The surrounding rocks of the deposits
are distinct, with the roof consisting of the Heshan Formation lime-
stone or iron-aluminum clayey, aluminum-rich claystone, or carbo-
naceous mudstone, and the floor is composed of the Heshan
Formation iron-aluminum rock or Maokou Formation limestone
(Liao and Wei, 2022). The Dajia mine is located in Xinxu Town,
Jingxi City. Figure 1b shows the distribution of surface geologic for-
mations in Dajia mining area. The bauxite layers are mainly inter-
bedded within Heshan Formation limestone. The area is rich in faults.
Due to high iron and sulfur content in the mineral composition,
the sedimentary bauxite deposits in the western Guangxi Province
have not been extensively exploited by the industry (Wang, 2016).
However, with the rapid development of technology, there is an in-
creasing demand for bauxite. The salento-type bauxite deposits in
Guangxi are nearing depletion, and it becomes quite necessary to
explore a deep sedimentary bauxite deposit.
DATA ACQUISITION
Large-scale data acquisition strategy
A conventional TEM survey collects data within one-third to
one-half of the transmitting loop (Figure 2), in which the primary
field is relatively uniform, making it suitable for imaging and
anomaly detection. However, deploying the transmitting loop is
the most time-consuming part of TEM fieldwork. Meanwhile,
the area for conventional data collection in each loop is small,
which results in low efficiency in data acquisition. In this study,
an efficient TEM data acquisition method was adopted that in-
volves data acquisition within the entire loop and at a certain range
outside the loop. From Figure 2, we can see that the new data
collection strategy collects EM data on 100 sites for one single
transmitting loop, which is 25 times the area covered by the con-
ventional acquisition method.
Survey layout
We acquired the data in the Dengli-Tianyanggumei mine of De-
bao County (Figure 3a)andDajiamineofJingxiCity(Figure3b),
Guangxi Province. For both working areas, we use the WGS84
coordinate system with Gaussian projection. The central meri-
dians are 108° (Figure 3a) and 105° (Figure 3b), respectively.
The transmitter coil has a size of 120 m ×120 m, transmitting
a square waveform with a base frequency of 50 Hz and a peak
current of 16 A. In the Dengli-Tianyanggumei working area, we
set up three transmitters and ten survey lines, with a 20 m interval
for line and point spacing, producing a total of 184 survey points.
Figure 1. Regional geologic maps. (a) Topo-
graphic map of Guangxi Province, China and
(b and c) the mining areas of Dengli-Tianyanggu-
mei and Dajia, respectively.
Figure 2. Comparison between our new TEM data acquisition
strategy and the conventional one.
Imaging of bauxite using TEM B133
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
The survey lines run from northwest to southeast. The Dajia
working area has five transmitters and 19 survey lines aligned
in northeast direction, with a total of 584 survey points. The
T4 transmitter developed by the Phoenix Company in Canada
is used to transmit the current, and the V8 Multifunction Receiver
is used to record the EM signal. The recording time ranges from
8.57 ×10−5to 7.35 ×10−3s, with a total of 20 time channels in
logarithmically equidistant.
DATA PROCESSING AND INVERSION
Preprocessing
To better analyze the data acquired by the TEM data acquisition
system adopted in this paper, we carried out a forward modeling for
the same acquisition strategy based on a finite-element method with
unstructured tetrahedral grids. Figure 4a–4h shows TEM responses at
different receivers for the data acquisition method in Figure 3for a
Figure 3. Layout of TEM survey. (a) Dengli-Tia-
nyanggumei survey area and (b) Dajia survey area.
There are three known boreholes in two working
areas, namely ZK0391 and ZK0392 in Dengli-Tia-
nyanggumei, and ZK104A01 in Dajia. The survey
points used in Figure 4are marked with purple text.
Figure 4. The TEM response of homogeneous
subsurface model with real topography and field
TEM response. (a–p) The measured TEM response
at the sites marked in Figure 3,(a–h) the TEM re-
sponse of the homogeneous subsurface model, and
(i–p) field TEM response.
B134 Wei et al.
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
homogeneous subsurface with topography. It is evident that the data
collected within the transmitting coil exhibit a monotonically de-
creasing attenuation curve, whereas the data outside the transmitting
coil experience a signal polarity reversal in the early traces, followed
by a monotonic decrease. The reason for this phenomenon is as fol-
lows: when the transmitting current is switched off, as the induced
underground eddy currents are similar in magnitude to the transmit-
ting coil, an opposite response inside and outside the coil is induced.
As the underground eddy currents diffuse deeper, their radius gradu-
ally increases, causing all measurement points to be located within
their confines of influence during the middle and late stages, thus
maintaining consistent polarities. From Figure 4i–4p, one can see that
the responses from randomly selected survey points conform physi-
cally to the diffusion law, exhibiting smooth and descending curves.
This also validates the good quality of our survey data.
During the field data acquisition process, external interferences are
encountered. To address this, we use methods such as stacking, trace
extraction, and outlier removal to preprocess the raw data. Later figures
(see the “Dengli-Tianyanggumei mine in Debao County”and “Dajia
mine in Jingxi city”subsections) show the fits of the predicted and
field data at different sites. It can be seen that the data fits are excellent.
3D inversions
Considering that 3D inversion of TEM data is a classic ill-posed
problem, we define an objective function including a regularization
term (Tikhonov and Arsenin, 1977;Constable et al., 1987) defined as
ϕðm;dpreÞ¼kWdðdobs −dpreÞk2
2þλkWmðm−m0Þk2
2;(1)
where Wdis a diagonal matrix with elements representing the recip-
rocals of the standard deviation of the noise, Wmis a model rough-
ness matrix, dobs and dpre denote the observed and predicted data,
respectively, and mand m0represent the estimated and reference
models, respectively. To achieve precise discretization of complex
Figure 6. The TEM inversion of Dengli-Tia-
nyanggumei mining area and data fit. (a) Rms ver-
sus iterations and (b) data fit of all points. The red
line indicates the position of high-voltage power-
line. (c–f) The preprocessed TEM responses (with
error bars) and computed data from the con-
structed model at different locations in (b).
Figure 5. A uniform half-space model used for TEM inversion of
Dengli-Tianyanggumei mining area. (a) Schematic diagram of sur-
vey points after coordinate transformation and (b) model discreti-
zation using unstructured tetrahedral grids. The black rectangles
and points in plot (a) indicate the transmitting loops and the survey
points, respectively.
Imaging of bauxite using TEM B135
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
models and loop transmitters, we use unstructured tetrahedral grids
that can effectively model the undulating terrain and complex struc-
tures in the underground. Taking partial derivatives of both sides of
equation 1with respect to the model parameters mn−1,weobtainthe
gradient of the objective function at the nth iteration as follows:
gn¼−2JTWT
dWdrn−1þ2λWT
mWmðmn−1−m0Þ;(2)
where rn−1¼dobs−dn−1
pre and Jis the sensitivity matrix that describes
the degree to which changes in model parameters affect the observed
data. The model update Δmis obtained via the L-BFGS method
(Liu and Nocedal, 1989). The new model is then obtained by posing
mnþ1¼mnþΔm.
To reduce the computational costs, we use the adjoint state
method to calculate the gradient of the objective function in equa-
tion 2, which requires only two forward calculations per iteration.
For more details about the adjoint state method, please refer to
Plessix (2006) and Liu et al. (2019). Since each forward calculation
involves a separate calculation for all transmitters, the computa-
tional cost of inversion is linearly proportional to the number of
transmitters. We adopt a large-scale survey configuration (see Fig-
ure 2) that significantly reduces the number of transmitters and con-
sequently reduces the cost for our 3D inversions.
INVERSION AND INTERPRETATION
OF TEM SURVEY DATA
Dengli-Tianyanggumei mine in Debao County
To facilitate the 3D meshing of the topography in the survey area,
we apply a coordinate transform to convert the geodetic coordinate
system to a Cartesian coordinate system (Figure 5a).Thesizeofthe
model in the survey area was set to 400 m ×600 m ×400 m. We
divide the model into 463,751 tetrahedral elements. As shown in Fig-
ure 5b, the unstructured tetrahedral mesh enables a flexible charac-
terization of the complex terrain. The meshes at the transmitters and
receivers are also refined to guarantee the accuracy of forward mod-
eling. Since the surrounding rocks in the survey area are predomi-
nantly limestone with relatively higher resistivity, a background
resistivity of 500 Ω·m was assumed. The resistivity of air was set
to 1 ×108Ω·m. For the inversion, we set the noise as 5% of the
data amplitudes (Qi et al., 2021;Cheng et al., 2023). Considering
that some survey data have extremely small values, particularly in
areas where the EM signal changes signs, or at late-time channels
where the responses show significant noise contamination and large
errors (as shown in Figure 6), a fixed noise floor is introduced. Add-
ing a noise floor ensures that the small signal, which is generally
overwhelmed by noise, does not impede the inversion process. In
this way, we can also stabilize the inversion as the influence of noisy
data at late-time channels is reduced (Wu et al., 2023).
The data processing and inversion in this study were conducted
on a workstation with 128 GB memory and 2.9 GHz CPU. The 3D
inversion of the Dengli-Tianyanggumei data takes 500 iterations
and 224 h. It also uses approximately 20 GB of memory. The root
mean square (rms) errors reduce from 9.31 to 1.83, yielding a stable
convergence (see Figure 6a). From Figure 6b, one can see that the
overall data fit is good, with the majority of survey points having an
rms value below 2.0. However, at survey points close to the power-
line, the data are contaminated, so the data fit worsens. However,
from Figure 6we can see that the inversion of the survey data shows
a good convergence.
Figure 7. The 3D inversions of TEM data from
Dengli-Tianyanggumei mining area. (a) Holistic
3D inversion result, (b) iso-surface of anomaly be-
tween 0.3 and 3 Ω·m, and (c and d) resistivity pro-
files cross the boreholes ZK0391 and ZK0392,
respectively.
B136 Wei et al.
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
Figure 7a shows the 3D inversion results of the Dengli-Tianyang-
gumei TEM data. We can see that the exploration method used in
this study effectively delineates the electrical layers of the subsur-
face. As shown in Figure 7b, the spatial distribution of low-resis-
tivity anomalies in the range of 0.3 to 3 Ω·m exhibits a tilted slab-
like pattern, suggesting a combination of sedimentary bauxite ore
layers and upper/lower conductive layers. By combining different
directional slices, we can infer that the bauxite ore body extends
from the southeast to the northwest with an increasing depth.
The ore layer appears relatively continuous. The resistivity profiles
obtained from the drilling holes ZK0931 (Figure 7c) and ZK0392
(Figure 7d) intersect with the low-resistivity layer, revealing the
presence of the bauxite ore body. Further analysis of the composi-
tion of the conductive layer can be conducted based on the drilling
samples obtained from these boreholes.
Tables 1and 2show the stratigraphic information for boreholes
ZK0931 and ZK0392. It can serve as a crucial reference for our
lithologic interpretation. Meanwhile, we have integrated the results
obtained from ERT profiling, shallow seismic reflection using ar-
tificial sources, and geologic information obtained from the Dengli-
Tianyanggumei mining area into a comprehensive interpretation.
From Figure 8b, it is seen that the geologic structure in the Den-
gli-Tianyanggumei mining area exhibits no significant alterations
or instability, with consistent rock formations and structural integ-
rity, and is predominantly characterized by a three-layer formation
with high-low-high resistivities. The shallow high-resistivity layers
primarily consist of microcrystalline and argillaceous limestones,
while the underlying high-resistivity layers consist of limestone in-
terbedded with dolomite. Figure 8c shows geologic stratigraphic
information inferred from rock samples obtained from drilling. It
is seen that the bauxite layer has a thickness of approximately
Figure 8. Geologic cross section and 3D inversion
profile cross two boreholes in Dengli-Tianyanggu-
mei mining area. (a) ERT result, (b) 3D TEM in-
versions, (c) geologic section, and (d) shallow
seismic result. The blue and red lines represent
the inferred boundaries between the high-velocity
zone and the bedrock, respectively.
Table 1. Lithology in borehole ZK0391.
Layer
index Stratigraphy Depth (m) Lithology
1Q0–2.30 Quaternary sediment
2P
3h22.30–33.13 Interbedded mudstone and
microcrystalline limestone
3P
3h233.13–39.43 Argillaceous sandstone
limestone
4P
3h139.43–39.79 Carbonaceous mudstone
5P
3h139.79–42.02 Sedimentary bauxite
6P
2m42.02–52.99 Micrite
Imaging of bauxite using TEM B137
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
2 m. Due to the limited resolution of the TEM method, achieving
detailed images of such a thin layer is rarely possible. However,
since the carbonaceous mudstone and ferruginous mudstone with
a thickness of 2–4 m overlying the bauxite layer are both conduc-
tive, they can induce significant TEM anomalies, so that they are
recovered together as one conductive layer.
By overlapping the bauxite layer from Figure 8c on the 3D in-
version profile, one can clearly see that the position of the bauxite
layer aligns well with the low-resistivity one, confirming the accu-
racy of our 3D inversion results. Furthermore, it is observed that the
low-resistivity layer in the inversion results appears to be thicker
than the actual formation, which is attributed to the use of smooth
model constraints during the inversion. This can result in smeared
boundaries and smooth bands above and below the low-resistivity
layer. Compared with the TEM method, the ERT only provides re-
sistivity structures for depths of approximately 50 m (Figure 8a).
Only the bauxite layer near the surface on the right-side is recovered
from the ERT results, while the deeper mineral layers cannot be
detected. Figure 8d shows the distribution of high-velocity zones
inferred from the shallow seismic reflection that roughly delineates
the lithologic boundary between the bauxite layer and the underly-
ing bedrock of limestone. However, due to the limited energy of the
artificial seismic source, an effective delineation of the deeper rock
layers is not achieved.
By integrating the results from multiple geophysical exploration
methods, it is inferred that the sedimentary bauxite deposits in the
Dengli-Tianyanggumei mining area exhibit a relatively continuous
and layered distribution, extending from southeast to northwest in a
shallow-to-deep manner. The bottom of the bauxite layer is char-
acterized by high-velocity and high-resistivity limestone. The dis-
tribution of formations in the survey area appears to be continuous.
The results from our TEM exploration and interpretation align well
with the geologic information.
To analyze the physical processes involved in TEM detection of
bauxite deposits, we calculate the current density at a vertical profile
x= 100 m for the inverted model. As shown in Figure 9, in the early
stages, the current density is primarily concentrated near the sur-
face, and the lower boundary aligns well with the lower boundary
of the ore body (Figure 9a). As time passes, the “smoke ring”of
current density moves downward and outward, with the near-sur-
face current density weakening (Figure 9b). At the remaining meas-
urement time, the highest current density remains within the ore
body and does not propagate to the formations below it (Figure 9c
and 9d). This leads to the limited resolution of TEM method for
detecting the area below the bauxite deposits.
Dajia mine in Jingxi city
We apply a similar coordinate transformation to align the survey
lines with the Cartesian coordinate system for the Dajia working
area (Figure 10a). The size of the central region of the discretized
model is 560 m ×560 m ×300 m. We use a fine grid to adapt to the
complex terrain (Figure 10b). The entire inversion model consists of
545,172 unstructured tetrahedral elements. The air resistivity is set
to 108Ω·m, whereas the background resistivity is set to 500 Ω·m.
The noise estimation remains consistent with that used in the Den-
gli-Tianyanggumei working area. The 3D inversion of Dajia TEM
data takes 500 iterations for a total of 426 h. It uses approximately
30 GB of memory. The rms decreases from 10.7 to 3.4, showing a
stable convergence (Figure 11a). From the data fit in Figure 11b,we
can see that the presence of surface pipes and powerlines cause a
poor data fit. However, the other points in this area exhibit good data
fit, with rms values around one. Thus, we conclude that the TEM
inversion in this area is successful.
Within the Dajia working area, borehole ZK104A01 exists,
whose location is provided in Figure 12. We can analyze the reli-
Figure 9. Current density distribution in Dengli-Tianyanggumei min-
ing area (x= 100 m). (a) t= 8.573 ×10−5s, (b) t= 1.584 ×10−4s,
(c) t=5.409×10−4s, and (d) t=1.0×10−3s. The arrows represent
the current direction.
Table 2. Lithology in borehole ZK0392.
Layer
index Stratigraphy Depth (m) Lithology
1Q0–6.70 Quaternary sediment
2T
1m16.70–57.46 Micrite
3T
1m157.46–70.19 Muddy limestone mixed
with mudstone
4P
3h270.19–114.23 Muddy sandy limestone
mixed with carbonaceous
mudstone
5P
3h2114.23–152.92 Interbedding between
bioclastic limestone and
muddy sandy limestone
6P
3h1152.92–154.15 Argillaceous siliceous rock
7P
3h1154.15–157.77 Carbonaceous mudstone
8P
3h1157.77–159.45 Iron aluminum mudstone
9P
3h1159.45–161.61 Sedimentary bauxite
10 P2m161.61–170.60 Bioclastic limestone mixed
with dolomite and muddy
sandy limestone
B138 Wei et al.
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
ability of our 3D TEM inversion results based on the geologic in-
formation obtained from this borehole. Figure 11 shows that the
resistivity profiles across ZK104A01 in different directions exhibit
a three-layer structure, characterized by high-low-high resistivity
variations. The target conductive layer, ranging from 10 to
30 Ω·m, primarily occurs within a shallow depth range of 0–50 m.
Unlike in the Dengli-Tianyanggumei mining area, the conductive
layer here presents an intermittent and discontinuous distribution,
which can be attributed to the presence of faults and karst caves
within the survey area. By comparing the lithologic information
of borehole ZK104A01 in Table 3, we see a good correlation be-
tween the depth of the low-resistivity layer obtained from our 3D
inversion and the actual occurrence of bauxite deposits. This im-
plies that our inversion results are reliable. In Figure 12b, we in-
ferred the locations of faults that significantly influence the
occurrence of bauxite deposits.
The formation of faults is closely related to historical geologic
events, which in turn can cause variations in the location and shape
of bauxite deposits. By integrating known faults and geophysical
inversions, we can predict the spatial distribution and shape of baux-
ite ores. To understand the distribution of controlling structures such
Figure 11. The TEM inversion of Dajia mining
area and data fit. (a) Rms versus iterations and
(b) data fit of all points. The red line indicates
the position of high-voltage powerline, whereas
the green one marks the location of metal pipeline.
(c–f) The preprocessed TEM responses (with er-
ror bars) and computed data from the constructed
model at different locations in (b).
Figure 10. A uniform half-space model used in TEM inversion of
Dajia mining area. (a) Planar view of survey points after coordinate
transformation and (b) model discretization with tetrahedral grids.
The black rectangles and points in (a) indicate the transmitting loops
and the measuring points, respectively.
Table 3. Lithology in borehole ZK104A01.
Layer
index Stratigraphy Depth (m) Lithology
1Q0–2.10 Quaternary sediment
2P
3h2.10–29.87 Containing cherty nodular
limestone
3P
3h29.87–30.62 Marl
4P
3h30.62–31.83 Carbonaceous marl
5P
3h31.83–33.94 Bauxite rock and bauxite
6P
2m33.94–43.65 Bioclastic limestone
Imaging of bauxite using TEM B139
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
as faults, we conducted several ERT profiles within the Dajia work-
ing area to obtain detailed images of shallow geologic structures
(specific locations of the survey line can be seen from Figure 13a).
From Figure 13b,13d, and 13f, it is seen that the ERT profiles can
clearly locate the near-surface faults. The inferred positions and dips
of the faults align well with TEM inversion results (Figure 13c,13e,
and 13g). By combining the inversion results from ERT and TEM, it
is clear that the presence of faults significantly influences the
occurrence of bauxite deposits. The bauxite layers no longer exhibit
continuous distribution but rather appear fragmented and seg-
mented. In Figure 14, we infer the planar distribution of surface
and subsurface faults at the elevation of 835 m based on the known
faults near the mining area and the results of this survey. The planar
distribution in Figure 14c clearly delineates the regions within the
underground bauxite zone. The upper-right area can be inferred as
the enriched deposit at this depth.
Figure 13. Structural controls inferred from TEM
and ERT inversion results. (a) The 3D TEM inver-
sion results, (b, d, and f) ERT results of lines L1,
L10, and L19, and (c, e, and g) the corresponding
TEM inversion results.
Figure 12. The 3D inversions of TEM data across the borehole ZK104A01 in Dajia mining area. (a) Resistivity profiles cross ZK104A01 and
(b) resistivity profile along surveying line L16 (marked by red line).
B140 Wei et al.
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
CONCLUSION
In this study, we have successfully used new TEM observations and
3D inversion techniques for the exploration of sedimentary bauxite de-
posits in western Guangxi Province. Our approach can simultaneously
acquire in-loop and out-of-loop TEM data and has greatly enhanced
the efficiency of data acquisition. By using unstructured tetrahedral
grids for the inversion, we obtain high-resolution models. Thanks to
the explorations conducted in the Dengli-Tianyanggumei and Dajia
mining areas, we can confirm that our new TEM observation and in-
version techniques can image the spatial distribution of bauxites effi-
ciently and accurately. The recovered low-resistivity zones correspond
to the known bauxite layers well. This provides valuable guidance for
subsequent mining operations. In conjunction with the geologic struc-
tures obtained from ERT surveys, we also successfully inferred the
distribution of controlling faults in the Dajia mining area, which serve
as crucial information for inferring the distribution of bauxite deposits.
ACKNOWLEDGMENTS
The paper was financially supported by the Guangxi Science and
Technology Program project (grant nos. GuikeAB21196028) and
the National Natural Science Foundation of China (grant nos.
42030806, 42074120, 41904104, 42174167, and 42004125).
DATA AND MATERIALS AVAILABILITY
Data associated with this research are confidential and cannot be
released.
REFERENCES
Ayman, N. Q., G. A. Abdulrahman, K. H. Ibraheem, M.
Khyzer, and M. A. Mazen, 2015, Performing high
resolution seismic reflection for mapping bauxite
layers: ASEG Extended Abstracts.
Bi, B., 2009, The application of the electric method to
bauxite exploration: Geophysical and Geochemical
Exploration, 33, 400–402.
Bogatyrev, B. A., and V. V. Zhukov, 2009, Bauxite
provinces of the world: Geology of Ore Deposits,
51, 339–355, doi: 10.1134/S1075701509050018.
Cablik, V., 2007, Characterization and applications of
red mud from bauxite processing: Gospodarka Sur-
owcami Mineralnymi-Mineral Resources Manage-
ment, 23,27–38.
Cheng, M., D. K. Yang, and Q. Luo, 2023, Interpreting
surface large-loop time-domain electromagnetic data
for deep mineral exploration using 3D forward mod-
eling and inversion: Minerals, 13, 34, doi: 10.3390/
min13010034.
Constable, S. C., R. L. Parker, and C. G. Constable,
1987, Occams inversion: A practical algorithm for
generating smooth models from electromagnetic
sounding data: Geophysics, 52, 289–300, doi: 10
.1190/1.1442303.
Dentith, M., and S. T. Mudge, 2014, Geophysics for the
mineral exploration geoscientist: Cambridge Univer-
sity Press.
Dragičević, I., M. Andrić, and I. Blašković, 1991, Geo-
logical-geophysical exploration of the bauxite de-
posits application of the shallow seismic reflection
method: The Mining-Geological-Petroleum Engi-
neering Bulletin, 3,23–28.
Gawu, S., E. Amissah, and J. Kuma, 2012, The pro-
posed alumina industry and how to mitigate against
the red mud footprint in Ghana: Journal of Urban and
Environmental Engineering, 6,48–56, doi: 10.4090/
juee.2012.v6n2.048056.
Guo, H., and Y. Lu, 2015, The application of audio
magnetotelluric sounding to Tuanxi bauxite deposit exploration: Yunnan
Geology, 34, 438–443.
Guo, Z., G. Xue, J. Liu, and X. Wu, 2020, Electromagnetic methods for
mineral exploration in China: A review: Ore Geology Reviews, 118,
103357, doi: 10.1016/j.oregeorev.2020.103357.
Hao, S., 2022, Research on the Application of CSAMT in bauxite explora-
tion (in Chinese): Huabei Natural Resources, 6,50–53.
Huang, G., J. Huang, S. Li, M. Wei, and Y. Tao, 2018, The characteristics
of mineral deposits and high density resistivity applied to detect
sedimentary bauxite of Western Guangxi: Mineral Resources and
Geology, 32,321–326.
Lai, Y., Z. Xi, F. Zhang, Y. Shi, H. Li, and Y. Xiang, 2021, Application of
opposing coils transient electromagnetic resistivity spectrum method to
detect bauxite deposits: Journal of Central South University (Science
and Technology), 52, 3264–3272.
Li, H., W. Askar, F. Li, Y. Jiao, Q. Wang, and Y. Zhou, 2013, Examination
study of geophysical methods in exploring buried bauxite deposit in
Southeast of Chongqing City: An example from the Chepan mining
area: Acta Geologica Sinica, 87, 384–392.
Li, T., J. Li, Q. Zhang, and X. Li, 2019, The dual application of transient
electromagnetic method in bauxite exploration and Goaf detection (in
Chinese): Resource Information and Engineering, 34,27–28.
Liao, J., and S. Wei, 2022, Analysis on the metallogenic law of sedimentary
bauxite in the Dengli Tianyang Gumei mining area of Debao County,
Guangxi (in Chinese): China Metal Bulletin, 2022, 246–248.
Liu, D. C., and J. Nocedal, 1989, On the limited memory BFGS method
for large scale optimization: Mathematical Programming, 45, 503–528,
doi: 10.1007/BF01589116.
Liu, X. F., Q. F. Wang, J. Deng, Q. Z. Zhang, S. L. Sun, and J. Y. Meng,
2010, Mineralogical and geochemical investigations of the Dajia Salento-
type bauxite deposits, Western Guangxi, China: Journal of Geochemical
Exploration, 105, 137–152, doi: 10.1016/j.gexplo.2010.04.012.
Liu, Y., C. Yin, C. Qiu, Z. Hui, B. Zhang, X. Ren, and A. Weng, 2019, 3-D
inversion of transient EM data with topography using unstructured
tetrahedral grids: Geophysical Journal International, 217, 301–318,
doi: 10.1093/gji/ggz014.
Maher, M. J., 1992, Transient electromagnetic surveys in the Okiep District:
Geophysics, 57, 736–744, doi: 10.1190/1.1443287.
Mongelli, G., R. Buccione, and R. Sinisi, 2015, Genesis of autochthonous and
allochthonous Apulian karst bauxites (Southern Italy): Climate constraints:
Sedimentary Geology, 325,168–176, doi: 10.1016/j.sedgeo.2015.06.005.
Figure 14. Inferred faults and distribution of bauxite deposits. (a) Known and inferred
faults in the working area and (b and c) the distribution of inferred faults on earth surface
and at the elevation of 835 m. The solid red lines indicate the unknown faults, whereas
the dashed lines indicate the inferred ones.
Imaging of bauxite using TEM B141
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
Nabighian, M. N., and J. C. Macnae, 1991, Electromagnetic methods in
applied geophysics: Volume 2, Application, Parts A and B: SEG.
Nogueira, P., M. P. Rocha, W. Borges, L. S. D. Cunha, E. X. Seimetz, M. M.
Cavalcanti, and P. A. D. Azevedo, 2011, Use of seismic refraction and
resistivity in bauxite deposit in the region of Barro Alto–Goiás, Brazil:
12th International Congress of the Brazilian Geophysical Society.
Orfanos, C., K. Leontarakis, K. Polychronopoulou, G. Apostolopoulos, K.
Athanasas, and N. Martakis, 2021, An integrated feasibility study using
passive geophysical methods for the investigation of the Gerolekas baux-
ite deposits site, Greece: Near Surface Geophysics, 20, 607–622, doi: 10
.1002/nsg.12181.
Plessix, R. E., 2006, A review of the adjoint-state method for computing
the gradient of a functional with geophysical applications: Geophysical
Journal International, 167,495–503, doi: 10.1111/j.1365-246X.2006
.02978.x.
Polychronopoulou, K., A. Lois, and D. Draganov, 2020, Body‐wave passive
seismic interferometry revisited: Mining exploration using the body waves
of local microearthquakes: Geophysical Prospecting, 68,232–253, doi: 10
.1111/1365-2478.12884.
Qi, Y., Q. Zhi, X. Li, X. Jing, Z. Qi, N. Sun, J. Zhou, and W. Liu, 2021,
Three-dimensional ground TEM inversion over a topographic earth con-
sidering ramp time: Chinese Journal of Geophysics, 64, 2566–2577.
Qin, F., X. Meng, and B. Yin, 2019, Distribution characteristics and ore-con-
trolling conditions of sedimentary bauxite deposits in Western Guangxi:
World Nonferrous Metals, 2019,92–94.
Tikhonov, A. N., and V. Y. Arsenin, 1977, Solutions of ill-posed problems:
SIAM Review, 21, 266–267.
Wang, K., K. Liu, and X. Zhang, 2010, Effect and analysis of searching for
bauxite deposit using transient electromagnetic method: Northwestern
Geology, 43,80–84.
Wang, W., 2016, The geological characteristics and material sources of
Permian bauxite in Western Guangxi: Master’s thesis, China University
of Geosciences.
Wei, G., 1999, Geology and ore-controlling factors of the accumulated baux-
ite deposit in Western Guangxi: Mineral Exploration, 8, 459–461.
Wei, M., J. Zhao, C. Yang, G. Liu, and S. Song, 2011, The application of
CSAMT to deep exploration in a bauxite ore district of Henan province:
Geophysical and Geochemical Exploration, 35, 600–603.
Wu, Y. Q., H. L. Xie, Y. J. Ji, P. Zhao, and Y. B. Wang, 2023, The potential of
the horizontal component TEM data in the detection of polarizable min-
eral: Synthetic cases: Minerals, 13, 523, doi: 10.3390/min13040523.
Xue, G., X. Li, and Q. Di, 2008, Research progress in TEM forward model-
ing and inversion calculation: Progress in Geophysics, 23, 1165–1172.
Xue, G., Z. Li, W. Guo, and J. Fan, 2020, The exploration of sedimentary
bauxite deposits using the reflection seismic method: A case study from
the Henan Province, China: Ore Geology Reviews, 127, 103832, doi: 10
.1016/j.oregeorev.2020.103832.
Xue, G., N. Zhou, W. Guo, J. Fan, K. Lei, and S. Zhang, 2022b, Delineation
of sedimentary bauxite deposits in Shaanxi Province using the gravity and
transient electromagnetic methods: Ore Geology Reviews, 144, 104865,
doi: 10.1016/j.oregeorev.2022.104865.
Xue, G., N. Zhou, W. Guo, and S. Zhang, 2022a, A metallogenic model for
bauxite deposits and geophysical prospecting methods: Using the sedi-
mentary type in Northern China as an example: Frontiers in Earth Science,
10, 791250, doi: 10.3389/feart.2022.841456.
Yang, R., G. Xu, L. Chu, Z. Cai, and B. Liu, 2019, Gravity-CSAMT inte-
grated metallogenic prediction of the bauxite deposit in Yushan-Liz-
huangzhai area, Henan Province: Metal Mine, 519, 113–122.
Zhang, B., K. W. Engebretsen, G. Fiandaca, H. Cai, and E. Auken, 2021, 3D
inversion of time-domain electromagnetic data using finite elements and a
triple mesh formulation: Geophysics, 86, no. 3, E257–E267, doi: 10.1190/
geo2020-0079.1.
Zhang, L., 2007, Application of transient electromagnetic method in explo-
ration of bauxite deposits: Geology and Prospecting, 43,68–71.
Zhang, Q., 2011, Metallogenic model and exploration techniques of the
bauxite, Western Guangxi, China: Ph.D. thesis, China University of
Geosciences.
Zhang, X., 2005, Prospecting of mid-deep bauxite resource by the transient
electromagnetic method: Chinese Journal of Engineering Geophysics, 2,
145–148.
Zhang, X., Y. Wang, and P. Chen, 2015, An analysis of the effect of the audio
magnetotelluric method (AMT) in the exploration of bauxite deposits in
Guizhou: Geophysical and Geochemical Exploration, 39, 283–287.
Biographies and photographs of the authors are not available.
B142 Wei et al.
Downloaded 03/30/25 to 59.72.97.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/page/policies/terms
DOI:10.1190/geo2023-0649.1
