Anda di halaman 1dari 19

Journal of Applied Geophysics 118 (2015) 4765

Contents lists available at ScienceDirect

Journal of Applied Geophysics

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

Magnetotelluric investigation of the geothermal anomaly in


Hailin, Mudanjiang, northeastern China
Lili Zhang a,b,, Tianyao Hao a, Qibin Xiao c, Jie Wang a, Liang Zhou d, Min Qi e, Xiangpan Cui a, Ningxiao Cai a
a
Key Laboratory of Petroleum Resources Research, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, People's Republic of China
b
State Key Laboratory of Petroleum Resource and Prospecting, China University of Petroleum, Beijing 102249, People's Republic of China
c
Institute of Geology, China Earthquake Administration, Beijing 100029, People's Republic of China
d
University of Utah, Salt Lake City 84112, USA
e
Institute of Mineral Resources Research, China Metallurgical Geology Bureau, Beijing 100025, People's Republic of China

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

Article history: To study the occurrence conditions and locations of geothermal bodies in Hailin, Mudanjiang, northeastern
Received 5 January 2014 China, we conducted a magnetotelluric investigation to delineate the electrical conductivity structure of the
Received in revised form 14 March 2015 area on three parallel proles. The area to the west of the Mudanjiang Fault lies in the Hailang sag of the Ning'an
Accepted 1 April 2015
Basin. The data were processed using the mutual reference technique, static shift correction, and structural strike
Available online 11 April 2015
and dimensionality analysis based on tensor decomposition. Moreover, a modied anisotropic-diffusion-based
Keywords:
method was used to suppress noise for the magnetotelluric time series data. This method retains the advantages
Magnetotelluric survey of conventional anisotropic diffusion and is superior in its discrimination ability. The method is characteristic not
Geothermal exploration only of the inherited features such as intra-region smoothing and edge preservation, but also of the adaptive se-
Modied-anisotropic-diffusion-based lection of the diffusion coefcient. Data analysis revealed that the electrical resistivity structure can be approxi-
denoising mated by a two-dimensional characterization. Two-dimensional inversion and rendering visualization show that
Hailang sag a highly resistive granite basement is covered with conductive sedimentary layers and that a relatively low-
The Mudanjiang Fault resistivity anomalous structure with a resistivity of approximately 100600 m is imbedded in the high-
resistivity background. The anomalous structure has a narrow top and a wide bottom (the bottom depth is at
least 3500 m). The shape and electrical features of the structure indicate favorable storage space for hot subsur-
face water. Fault activities and magma intrusion may result in the fractures of the basement, which are lled with
hot water and thus produce the relatively low resistivity. Based on a comprehensive analysis, we infer that the
structure is indicative of a geothermal reservoir. An exploratory well drilled near the structure conrms the oc-
currence of high temperatures. Several geological factors (cap rock, basement, and major faults) determine the
favorable geothermal conditions of the reservoir. Large areas of granite form the major thermal source for the
study area. The Mudanjiang and Hailang River Faults and their subsidiary faults provide another heat source
and movement channels.
2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction (especially hot water) and fractured zones usually have lower resistiv-
ity. The resultant low-resistivity anomalies have generally been the
High electrical conductivity is one feature of geothermal sources. main target for the geophysical exploration of geothermal resources
Geothermal-water-rich rocks commonly have relatively lower resistiv- (Simpson and Bahr, 2005). This is the geophysical basis of
ity than initial rocks, and the variation in the resistivity is related to the magnetotelluric (MT) soundings for geothermal resource exploration.
water abundance, temperature, and degree of mineralization (Spichak The MT method can be used to detect the distribution of highly conduc-
et al., 2007). Hot subsurface water is closely related to structural frac- tive geological bodies and faults. MT data, when combined with other
tured zones connected to medium-deep hot sources. Water-abundant geophysical anomalies and geological conditions, can be used to predict
the distribution of thermal reservoirs (Spichak and Manzella, 2009).
Since the late 1980s, the development of remote references, robust ap-
Corresponding author at: Key Laboratory of Petroleum Resources Research, Institute
of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, People's
proaches, distortion corrections, multidimensional modeling and inver-
Republic of China. sion methods has made the interpretation of magnetotelluric data more
E-mail address: lilyzhang@mail.iggcas.ac.cn (L. Zhang). reasonable (Bahr and Simpson, 2005; Berdichevsky and Dmitriev, 2008;

http://dx.doi.org/10.1016/j.jappgeo.2015.04.006
0926-9851/ 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
48 L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765

Chen and Wang, 1990; Telford et al., 1990). There have been many suc- After the MT survey, an exploratory well was drilled where the MT results
cessful applications of the MT method in geothermal exploration (Bai indicate favorable structures of geothermal water. The well conrms the
et al., 2001; Corwin and Hoover, 1979; Johnston et al., 1992; Kong occurrence of high temperatures. The geothermal sources are deduced
et al., 1991; Newman et al., 2008; Pellerin et al., 1996; Volpi et al., based on the geological conditions of the area.
2003; Wright et al., 1985).
Previous geophysical and geological studies suggest the presence 2. Geological and tectonic setting
of geothermal resources underneath the southern part of Mudan-
jiang and its adjacent areas in Heilongjiang Province, northeastern 2.1. Tectonic setting
China (Qi, 2006; Sui et al., 2011; Ye, 2011; Wang and Huang, 2014).
The activities of the Mudanjiang and DunhuaMishan Faults and The study area (see the red rectangle in Fig. 1) is located in the
their subsidiary faults have provided heat and favorable space for volcanic rock belt of Changbaishan Mountain, which is an important
the geothermal resources. Hailin lies in western Mudanjiang and is part of the Pacic Mesozoic volcanic chains. The area is next to the west-
close to the Mudanjiang Fault. Neither data nor cases of geophysical ern side of the Mudanjiang Fault, in the Hailang sag of the Hailang
investigations in Hailin have been published. Therefore, no geophys- Depression (belonging to the Ning'an Basin). The Hailang Depression
ical evidence can be referenced to reveal whether Hailin also has is a relatively independent artesian basin (Li et al., 2007). Strata are
geothermal potential like the southern part of Mudanjiang. Some widely outcropped, and intense magmatic activities and structures
geological and hydrogeochemical investigations reveal that the developed there. The structural evolution includes the Zhangguangcai
basement of Hailin mainly consists of granites and granodiorites Range epoch, Caledonian epoch, Indo-Chinese epoch, and Yanshan
and is buried at shallow depths (Jing et al., 2008; Li et al., 2007). As epoch. The deformation structures are mainly northsouth linear
a result, the basement is highly resistive. However, if the granites compaction folds.
or granodiorites are fractured or lled with uids, especially hot
water, their resistivity will decrease dramatically. Such a resistivity 2.2. Stratigraphy
contrast is to be expected for an MT investigation.
To study the possible occurrence conditions of Hailin, we conduct Fig. 2 is the regional geological map of the MT survey area. The out-
an MT survey on three parallel proles at the village of Hailin to delineate crop strata are distributed in valleys and are composed of sandstones
the geoelectrical structure and provide geophysical evidence for the and glutenites, which mostly belong to the Neogene and Quaternary.
exploitation and evaluation of geothermal resources in the area. Reliable Other outcrops close to the survey area are Cretaceous Hailang forma-
data are obtained through mutual reference, noise suppression, static tion, migmatitic granodiorite of Variscan age, granite rock formation of
shift correction, and dimensionality analysis based on tensor Early Yanshanian age, alaskite granite rock formation of Late Variscan
decomposition. A modied anisotropic-diffusion-based method is age, and granite porphyry rock formation of Late Yanshanian age. Be-
utilized to suppress noise for the MT time series. This method retains cause of the intensive magmatic activity in the area, the strata are seri-
the advantages of conventional anisotropic diffusion and is superior in ously damaged and the remains are mainly xenoliths distributed in the
its discrimination ability. A 2-D inversion is used to provide better con- intrusive rocks. The intrusive rocks are alaskite granite, granodiorite,
straints on the conductivity structures of the three proles. A rendering gabbro, and granite porphyry of Yanshanian epoch, Indo-Chinese
visualization of the inversion results is applied to represent the structures. epoch, Late Variscan epoch, and Proterozoic. Their modes of occurrence

Fig. 1. Topographic map.


L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765 49

Fig. 2. Regional geological map of the MT survey area (the coordinate system is WGS84, UTM (Universal Transverse Mercator) projection). (1Quaternary system; 2Hailang formation;
3Alaskite granite rock formation of Early Yanshanian Age; 4Alaskite granite rock formation of Late Variscan age; 5Granite porphyry; 6Proterozoic Migmatitic granodiorite;
7Migmatitic granodiorite of Variscan age; 8Daotaiqiao formation (the Neocene) basalt; 9Granite porphyry rock formation of Late Yanshanian Age; 10Granite rock formation of
Early Yanshanian Age; 11Aotou formation; 12Lower Aotou formation; 13Upper Aotou formation; 14Ningyuancun formation; 15Ximashan formation; 16MT sites.).

are abyssolith, dyke, and shoot. The geotectonic units control the devel- Along the contact between the strata and basement granites, the
opment of the regional stratigraphy, especially the pre-Mesozoic strata. presence of detrition belts (weathering fracture zones) helped to form
The detailed strata in Hailin are shown in Table 1. preferable migration pathways for heat and water.
After the DunhuaMishan and Mudanjiang Fault Belts (Fig. 1)
formed, the Hailang sag formed in the place where sandstones and silt-
stones were deposited, which has a maximum thickness greater than 2.3. Hydrogeological characteristics
1200 m. In Table 1, the sedimentary cover is mainly from the Cretaceous
Hailang Group's water-bearing rock formation, and the Eogene There are three types of underground water in Hailin. They are pore
Huanghua Group's water-bearing rock formation, Quaternary soft water in loose rocks, ssure-pore water in clastic rocks, and ssure
deposits and water-bearing basalt rock formation are sparsely distribut- water in bedrocks. The pore water in loose rocks is distributed in the
ed (Li et al., 2007). The lithology is mainly sandstone, siltstone, clay, and Eogene Huanghua Group's water-bearing rock formation, Quaternary
argillite. The thermal conductivity of the clay stone, siltstone, argilla- soft deposits and water-bearing rock formation. The water-bearing for-
ceous siltstone, and ne sandstone is 1.98 TCU, 4.76 TCU, 5.00 TCU, mation for the ssure-pore water in clastic rocks is the Cretaceous
and 3.98 TCU, respectively. The thermal conductivity values of the Hailang Group which consists mainly of sandstone, glutenite, and silt-
cover rocks are smaller than those of common sedimentary rocks. stone, interbedded with thin layers of mudstone. The ssure water in
Therefore, these sediments form a favorable cap for the Earth's heat bedrocks can be divided into weathering ssure water and structural
and provide good geologic conditions for geothermal sources. ssure water. The water-bearing formation of the former water is of
The basement of the Hailang sag is mainly Yanshanian epoch, the granites in different periods and the Paleozoic metamorphic rocks.
Variscan epoch and Lower Proterozoic granites and granodiorites (Jing The latter water exists in fault belts, the composite parts between differ-
et al., 2008). Its uranium content is (2.813.38) 106, with a maxi- ent structures, and the contact zones between different lithologies,
mum of (1015) 105, and its leaching rate is 30%90% (Li et al., where water-rich zones may form. The reciprocation of the under-
2007), indicating an abundance of uranium resources. The basement ground water is relatively strong. Basically, the ssure water in bedrocks
has complicated morphology. Its burial depth is approximately 100 ows through the soft deposits and the ssures of bedrocks inside the
400 m, with the maximum depth being 1200 m (Jing et al., 2008). depression (Jing et al., 2008).

Table 1
Strata in Hailin.

Stratum Lithological composition

Cenozoic Quaternary Alluvia, diluvial, sand clay, clayey soil, loess-like soil, bray stone. Along the Mudanjiang River
there distributed Jingbo-age basalt, less than 40 m in thickness.
Neogene Pliocene Daotaiqiao formation Unconsolidated glutenite and basalt
Eogene Eocene Huanghua formation Loose glutenite, argillite, sandstone mingled with argillite
Mesozoic Cretaceous Lower Hailang formation Sandstone, glutenite interbedded with thin layers of mudstone, siltstone
Houshigou formation Arcose, siltstone. Fractures and intermediate acidity dykite are developed in the formation.
Aotou formation Volcaniclastic rock, sedimentary rock, granite porphyry
Jurassic Upper Ningyuancun formation Acidic lava mingled with a little intermediate acidity lava and clasolite. Granite and granodiorite
Neopaleozoic Permian Granite, granodiorite, and alaskite granite
Lower Proterozoic Mashan Group Ximashan formation Granite and granodiorite, overlying alaskite granite rock formation of the late Variscan age
50 L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765

2.4. Major faults The Mudanjiang Fault Belt formed almost instantaneously with the
DunhuaMishan Fault Belt. Many synchronous subsidiary faults formed
The DunhuaMishan and Mudanjiang Faults (Fig. 1) are lithospheric with the two faults, which mostly trend northwest or northeast. The
faults in the Mudanjiang area. In the Late Jurassic and Early Cretaceous Mudanjiang Fault is a supracrustal fault between the Jiamusi and
approximately 150,000,000 years ago, the DunhuaMishan Fault Belt Songnen Blocks. It is a well-developed lithospheric fault cutting toward
formed in the eastern marginal active zones of the Eurasian plate. The the mantle and is a transportation channel for mantle uids into the
fault belt trends northeastsouthwest and is 10 to 20 km in width. crust (Zhang et al., 1998).

 p
Fig. 3. (a) A signal ux x2x=20 sin 4 x ; x0; 4000: (b) The noise-added signal. The signal (a) is added with speckle noise (evenly distributed random noise with variance of 0.0001).
(c) The noise suppression result of the signal (b) with the median lter. The MSE value between the original signal and the noise-suppressed result is 0.0729. (d) The noise suppression
result of the signal (b) with the PM model. The MSE value is 0.0175 when the iteration number is 20. (e) The noise suppression result of the signal (b) with the modied method. The MSE
value is 0.0169 when the iteration number is 20. (f) Spike noise. (g) The noise-added signal. The signal (a) is added with the spike noise (f). (h) The noise suppression result of the signal
(g) with the median lter. The MSE value is 0.0638. (i) The noise suppression result of the signal (g) with the PM model. The MSE value is 0.1405 when the iteration number is 20. (j) The
noise suppression result of the signal (g) with the modied method. The MSE value is 0.1027 when the iteration number is 20. (k) The noise-added signal. The signal (a) is added with the
sine function u(x) = 3 sin(100 x). (l) The noise suppression result of the signal (k) with the median lter. The MSE value is 3.3920. (m) The noise suppression result of the signal (k) with
the PM model. The MSE value is 0.0806 when the iteration number is 20. (n) The noise suppression result of the signal (k) with the modied method. The MSE value is 0.0746 when the
iteration number is 20.
L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765 51

Fig. 3 (continued).

In addition to the two fault belts, Hailin is near the westeast move and for heat to transfer, forming a source of subsurface heat.
trending Hailang River Fault and three other major faults, which trend Ning'an (see the location of Ning'an in Fig. 1) is in the southern
northeast, northnortheast, and northwest (Fig. 1). The Hailang River Mudanjiang area. In this location, the Jingbo Lake graben is close to
Fault spreads almost along the Hailang River (the location of the Hailang the intersection of the Mudanjiang and DunhuaMishan Faults. Exploi-
River is shown in Fig. 2) and intersects the Mudanjiang Fault in tation and drilling results (Sui et al., 2011; Wang and Huang, 2014) in-
Mudanjiang where some subsidiary faults formed along with the dicate that the Jingbo Lake graben can be a large-scale geothermal
Hailang River Fault. eld. It is deduced that the DunhuaMishan fault zone has been contin-
Under the northwestward subduction stress eld of the Pacic Plate, uously providing heat and water for the geothermal eld (Sui et al.,
the faults in Mudanjiang were frequently active and caused multiple 2011). This fault facilitated basaltic magmatic exhalation from the Neo-
phases of magmatic activity and intrusion (Wan and Zhong, 1997). In gene to the Quaternary. The DunhuaMishan Fault controls the Neo-
each phase of magmatic activity, magma moved along the faults in gene basalts, and its subsidiary faults control the Quaternary basalts
different directions and periods, either intruding into the Earth's surface (Wan and Zhong, 1997). The faults that dominated the Quaternary ba-
or extruding out. In a word, the faults provided channels for magma to salt eruption are the main geothermal sources. The drilling data
52 L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765

Fig. 3 (continued).

indicated that the geothermal water found in the Jingbo Lake graben is data. A modied anisotropic-diffusion-based method was applied for
stored in the pores of the sandstones and glutenites or the macroscopic the MT time series data to suppress noise.
cracks (fractures) resulting from the fault structural activities (Sui et al., Anisotropic diffusion lters have been used to smooth 3-D seismic
2011). data (Fehmers and Hocker, 2003) and remove aliasing artifacts from
Our survey area (Figs. 1, 2) is chosen near the intersection of the aeromagnetic data (Smith and O'Connell, 2005). The features of conven-
northnortheast-trending fault and the Hailang River Fault. We use tional anisotropic diffusion (i.e., the PM model proposed by Perona and
MT characterization to reveal whether the intersection could provide fa- Malik, 1990) include intra-region smoothing and edge (abrupt varia-
vorable space for heat and water and whether Hailin has geothermal tions or boundaries in anomalies, usually those with relatively large gra-
potential, similar to the southern part of Mudanjiang. dients) preservation. The relevant equations are listed in Appendix A.
The PM model overshadows some commonly-used smoothing lters
3. Magnetotelluric data (for example, the median lter) due to its edge-preserving feature.
However, the model still has its deciencies: it sometimes cannot suc-
3.1. MT data acquisition cessfully deal with noise in small spaces or abrupt changes with large
gradients (Alvarez et al., 1992); the parameter k is not easy to select.
This survey includes 18 stations distributed along three parallel pro- Our modication includes dimensionality and the selection of
les (Line-A, Line-B, and Line-C), with an average station distance of diffusion coefcients. Conventional anisotropic diffusion is applied to
500 m. The locations of the MT stations are shown in Figs. 1 and 2. two-dimensional images, while the modied method is applied to
The MT data were recorded with ADU06 systems manufactured by the one-dimensional MT time series. We select the diffusion coefcient
Metronix Company of Germany. The survey lines, Line-A, Line-B, and adaptively; that is, a larger degree of diffusion is applied to segments
Line-C, trended nearly westeast. Horizontal electric (E) and magnetic with smaller gradients, while conventional diffusion is applied to
(H) eld components were recorded simultaneously in the magnetic those with larger gradients. Therefore, the deciencies of the PM
northsouth and eastwest directions at each station. The vertical com- model can be overcome. The equations are shown in Appendix B.
ponent of the magnetic eld was also recorded. The electric eld in the x Fig. 3 displays three types of noise (speckle, spike, and sinusoidal)
direction (Ex) was measured in the northsouth direction and the eld and corresponding noise suppression with the median lter, PM
in the y direction (Ey) was recorded in the eastwest direction. The dis- model, and modied method. The PM model and modied method
tance between the electrodes was approximately 100 m. The magnetic both effectively suppress the speckle noise, superior to the median lter
component Hx describes the northsouth direction, Hy the eastwest (Fig. 3ce). These two methods are inferior to the median lter when l-
direction, and Hz into the ground surface. During the recording, the tering the spike noise (Fig. 3f) as they result in slightly larger mean
systems were set as synchronous by GPS antennas so that the remote square error (MSE) values (derived between the original signal and
reference algorithm (Gamble et al., 1979) can be used to calculate the the noise-suppressed results), but the median lter seems to retain larg-
impedance tensors. Four frequency bands were measured at each er burrs than the other two methods (Fig. 3hj). Smaller burrs are
site: the HF band (sample frequency 40,960 Hz, frequency range retained in the results of these two methods; thus, relatively large
10,000500 Hz) ran 7 s by default; the LF1 band's (sample frequency MSE values occur. These two methods overshadow the median lter
4096 Hz, frequency range 765 Hz23 Hz) recording time was 15 min; when suppressing the sinusoidal noise (Fig. 3ln), though they fail to
and the LF2 (sample frequency 64 Hz, frequency range 13 Hz2.68 s) completely restore the original signal. The results show that both the
and LF3 bands (sample frequency 2 Hz, frequency range 2.68 s PM model and modied method can effectively suppress the noise.
90.45 s) ran more than 12 h. The modied method inherits the advantages of the PM model, includ-
ing intra-region smoothing and edge preservation, which makes these
3.2. MT data processing and analysis two methods better than the median lter. However, the modied
method is a notch above the PM model due to its adaptive selection of
3.2.1. Noise suppression based on modied anisotropic diffusion diffusion coefcients, which is reected in two aspects: it can discrimi-
Man-made noise can be a major obstacle to the effective application nate the signal and noise even where the signal curve (Fig. 3a) has small
of the MT method when studying of electrical structures (Fontes et al., variation; and it results in slightly smaller MSE value.
1988). The survey area was 200 m away from village settlements and Fig. 4ab shows the raw and processed time series segment of Site
was somewhat interfered by noise. The noise in the survey area includ- A3. Only the data in the LF2 and LF3 bands were processed. The large
ed spikes or other random noise. We performed noise identication and spikes clearly visible in Fig. 4a are ltered out in Fig. 4b, leaving the use-
the suppression of time series segments to improve the quality of the ful data untouched. The apparent resistivity and phase curves of Site A3
L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765 53

Fig. 4. (a) A segment of the measured (raw) time series at Site A3. (b) The data after the noise suppression with the modied method. (c) Apparent resistivity and phase curves of Site A3
before the noise suppression. (d) Apparent resistivity and phase curves of Site A3 after the noise suppression.

show improvements in their shape and amplitude after noise suppres- We did not set an additional distant site for the remote reference
sion (Fig. 4cd). Note that the noise suppression is not applied to the technique; instead, we applied mutual reference with the magnetic
whole time series data but to the segments that need noise suppression data of synchronous MT sites. Fig. 5ab shows the effects of the mutual
after our manual discerning. This is performed to avoid the loss of useful reference processing without or with noise suppression combined. The
information. noise suppression was only applied to the data in the LF2 and LF3 bands,
as in Fig. 4d. The data from Site A3 (Fig. 4c) were improved after the pro-
3.2.2. Impedance cessing techniques were applied. The improvement is especially obvi-
The time series data were primarily processed with Mapros (soft- ous at periods of approximately 0.110 s (the so-called dead band).
ware from Metronix Inc.), and mostly stable and high-quality estimates The mutual reference processing when combined with the noise sup-
of the impedances were yielded in a frequency band ranging from pression has a slightly better effect (Fig. 5b).
8000 Hz to 0.01 Hz. There are commonly 46 frequency points for The apparent resistivity and phase curves for the MT sites (except
each site. Sites A6 and C6) are shown in Fig. 6. Sites A6 and C6 were excluded
54 L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765

10000
Apparent Resistivity
XY 3.2.3. Dimensionality and strike direction analysis
YX Structural strike and dimensionality analysis is based on tensor de-
(Ohm-m) 1000
composition methods (Groom and Bailey, 1989; Swift, 1967). The skew-
100 ness (Swift, 1967) can be used to qualitatively evaluate whether
electrical structures below MT sites deviate from two-dimensionality.
10 The dimensionality analysis, which calculates the skewness of imped-
ance tensors, reveals that most of the Swift's skewness values are
1 below 0.3 (Fig. 7), which is a reection of two-dimensionality struc-
tures. Thus, the data can be approximated by 2-D characterization. The
80
Phase( )

skewness is prone to static effects. Generally, it is assumed that a 1-D


setting is dominant in the HF or LF1 frequency band. However, the
40 skewness values of Site B5 in the HF and LF1 frequency band are around
or larger than 0.5, and there exists a split between the xy- and yx-
0 components of the apparent resistivity (Fig. 6). This is indicative of a
10000 1000 100 10 1 0.1 0.01 static shift at the recorded frequencies.
Frequency(Hz) The phase sensitivity skewness values within the range 10 Hz
103 Hz are smaller than 0.3 at most sites, except for a few high frequen-
(a) cies at Site B5. The low values (less than 0.3) may indicate that the
electrical structures at the corresponding frequencies are approximately
2-D. The larger values may be related to the presence of static shifts aris-
10000
Apparent Resistivity

XY ing from near-surface inhomogeneities (geological noise).


YX
(Ohm-m)

1000 Tipper information is utilized from the measured vertical compo-


nent of the magnetic eld. The induction arrows for both real and imag-
100 inary parts are derived from the tipper (Schmucker, 1970). The plotted
real arrows are considered intuitively easier to understand and their
10 presentation is often preferred. The real arrows point away from regions
of enhanced conductivity, while the imaginary arrows change sign at a
1
period where the real parts are maximal. The induction arrows have
been meaningfully used or tailored for dimensionality analysis (Chen
80
Phase( )

et al., 2004). Note that some authors choose to rotate the real part by
180 to make it point toward conductors. Over a 2-D structure, the
40
real parts of the induction vectors are orthogonal to the geoelectric
strike and can be used to resolve the 90 ambiguity in the strike direc-
0
tions derived from tensor decomposition (Padilha et al., 2006).
10000 1000 100 10 1 0.1 0.01 The induction arrows of the stations along the three proles are plot-
Frequency(Hz) ted in Fig. 8. The data do not present signicant variations in the resistiv-
ity structures, and the arrows in the low frequency bands (below 1 Hz)
(b) are very small. The arrows in the frequency band (101 Hz) are large, in-
dicating a roughly northsouth strike direction. These arrows consis-
Fig. 5. Apparent resistivity and phase curves of Site A3 after the mutual reference and noise tently point along the prole direction. The arrows at the frequencies
suppression. (a) Only the mutual reference is performed. (b) Both the mutual reference
and the noise suppression are performed.
larger than 10 Hz mostly point northwest or westnorthwest. Fig. 8 in-
dicates that the geoelectric strike may be northeast, northnortheast, or
northsouth. As mentioned above, the real induction arrows should
from the following process and inversion due to their bad data quality. point away from good conductors in a two-dimensional setting. The ar-
The apparent resistivity and impedance phase curves of Line A and rows may hint at a conductive anomaly body beneath the MT sites.
Line B indicate that the xy and yx curves are similar in shape, generally From the behavior of the arrows, we can conclude that a 2-D interpreta-
of the K-type or HK-type. The survey area is characterized by a low- tion will be meaningful for the data.
high-low resistivity variation from shallow to deep formations, which Strike direction information is calculated based on Groom and Bailey
is a three-layer model. (GB) impedance tensor decomposition (Groom and Bahr, 1992;
The xy and yx apparent resistivity curves of Line A are congruent for Groom and Bailey, 1989; Jones and Groom, 1993; McNeice and Jones,
high frequencies and split for lower frequencies, suggesting that a 2-D 2001), as shown in Figs. 9 and 10. Rose diagrams are used to depict G
structure in the deeper formation is covered by a 1-D layered structure. B strike directions of all of the sites for whole frequencies (Fig. 9) and
The curves of Sites A3 and A5 split at a higher frequency of 1000 Hz, in- the following four bands (Fig. 10): 30010 Hz, 101 Hz, 10.1 Hz, and
dicating thinner 1-D structures at shallow depths. The apparent resistiv- 0.10.01 Hz. Two dominant trends appear in Fig. 9a, and there are
ity increases at a frequency of 100 Hz for Sites A1, A2, and A3 and at three rather clear trends in Fig. 9b and c. The calculations for the four
1000 Hz for A4 and A5. This indicates that the thickness of this shallow frequency bands of Prole Line-C (Fig. 10) also show that the major
low-resistivity layer seems to decrease from west to east. The data in the strike direction is not unique and thus is difcult to determine. Mean
spectra of 8000 Hz to 1000 Hz were seriously interfered by noise for strikes were computed for the four frequency bands and are shown in
Sites B1, B2, and B4, and they were omitted. An obvious static shift can Fig. 11. In the frequency band of 30010 Hz, the mean strike of the pro-
be seen for Site B5, and a slight shift can be seen for Site B3. The deeper les is northeast or eastnortheast (Fig. 11a). In the LF2 frequency band
formation at Sites C1, C2, and C3 indicates a 1-D electrical structure, (101 Hz) a signicant scatter occurs for Prole Line-B and the mean
while C4 and C5 indicate a 2-D structure. In Fig. 2, the location of Site strike is northeast, northnortheast, eastnortheast, or nearly west
C4 is close to the outcropped migmatitic granodiorite of Variscan age. east (Fig. 11b). The change of strike suggests that there may be some
The dead-band effect caused a general uncertainty for the data in the contamination by the dead band effect at these periods. There is little
spectra of a few Hz to 10 s for Site C4, which is reected as large error variation among the sites of Prole Line-A, with a mean strike of north
bars. northeast. The low-frequency bands (0.10.01 Hz and 10.1 Hz) show a
L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765 55
XY YX
10000 10000 10000
Apparent Resistivity

1000 1000
(Ohm-m)

1000
100 100 100
10 10 10
1 1 1
Phase( )

80 80 80
40 40 40
0 0 0
100001000 100 10 1 0.1 0.01 10000 1000 100 10 1 0.1 0.01 10000 1000 100 10 1 0.1 0.01
A1 A2 A3
10000 10000 10000
Apparent Resistivity

1000 1000 1000


(Ohm-m)

100 100 100


10 10 10
1 1 1
Phase( )

80 80 80
40 40 40
0 0 0
100001000 100 10 1 0.1 0.01 100001000 100 10 1 0.1 0.01 100001000 100 10 1 0.1 0.01
A4 A5 B1
10000 10000 10000
Apparent Resistivity

1000 1000 1000


(Ohm-m)

100 100 100


10 10 10
1 1 1
Phase( )

80 80 80
40 40 40
0 0 0
100001000 100 10 1 0.1 0.01 10000 1000 100 10 1 0.1 0.01 100001000 100 10 1 0.1 0.01
B2 B3 B4
10000 10000
Apparent Resistivity

10000
1000 1000
(Ohm-m)

1000
100 100 100
10 10 10
1 1 1
80
Phase( )

80 80
40 40 40
0 0 0
100001000 100 10 1 0.1 0.01 10000 1000 100 10 1 0.1 0.01 100001000 100 10 1 0.1 0.01
B5 B6 C1
10000 10000 10000
Apparent Resistivity

1000 1000 1000


(Ohm-m)

100 100 100


10 10 10
1 1 1
Phase( )

80 80 80
40 40 40
0 0 0
100001000 100 10 1 0.1 0.01 100001000 100 10 1 0.1 0.01 100001000 100 10 1 0.1 0.01
C2 C3 C4
10000 Frequency(Hz) Frequency(Hz)
Apparent Resistivity

1000
(Ohm-m)

100
10
1
Phase( )
80
40
0
10000 1000 100 10 1 0.1 0.01
C5
Frequency(Hz)
Fig. 6. Apparent resistivity and phase curves of the MT sites.
56 L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765

A1 A2 A3 A4 A5 B1 B2 B3 B4 B5 B6 C1 C2 C3 C4 C5

1000
Frequency(Hz)
100

10

0.1

0.01
0 0.5 0 0.5 0 0.5 0 0.5 0 0.5 0 0.5 0 0.5 0 0.5 0 0.5 0 0.5 0 0.5 0 0.5 0 0.5 0 0.5 0 0.5 0 0.5
Skewness

Fig. 7. Skewness of the impedance tensors as dened by Swift (the missing frequency points are the points that were deleted before the skewness calculation due to the bad quality of the
apparent resistivity and phase).

consistent strike of northeast or northnortheast (Fig. 11cd). As men- accounts for the smoothness of a magnetotelluric sounding and its in-
tioned in the previous section regarding the tectonic setting, the defor- sensitivity to sharp, small-scale contrasts favors least-structure models.
mation structures in the area are mainly northsouth linear compaction In the inversion of the MT data, the initial model was given a homoge-
folds. The area is close to the northnortheast- and northeast-trending neous half space with 100 m and incorporates the topography from
major faults. The mean strike arrows are generally parallel to these the three MT proles. The TE and TM mode phases and apparent resis-
major faults. According to the eld layout of each MT site, the Ex compo- tivities were inverted simultaneously with 100 iterations. After 30 iter-
nent is in the northsouth direction. We did not rotate the data accord- ations, the root mean square (RMS) error values between the models
ing to Figs. 9 and 10 due to the scattered strike directions for the three and original data were all smaller than 5%. The error oors were initially
proles. Instead, the data were rotated by approximately 40 based on given as 25%, and then the phase error oor was reduced to forcibly t
Fig. 11. After the rotation the xy component corresponds to the TE polar- the model to the phase angle (the nal value is 5%). Afterwards, the
ization mode in which electric currents ow along the faults, while the error oor for the apparent resistivity was lowered to 5% as well because
yx component corresponds to the TM mode with electric currents the phase is generally better estimated by the processing than by the ap-
owing normal to the faults. parent resistivity. Various regularization parameters ( from 1 to 30)
were tested, and was assigned as 1. Numerous experiments have to
3.2.4. Static shift correction be conducted to achieve a best-tting and reliable model. The tting sit-
Static shifts distorted the apparent resistivity data of Sites B5 and B3. uations between the data and model responses were analyzed for all of
Several techniques have been developed and tested during the previous the sites and frequencies. A reference well is approximately 2000 m
decades for static shift correction/removal (Jones, 1988; Meju, 1996; away from the survey area (see the blue star in Fig. 2) and reached
Ogawa and Uchida, 1996; Pellerin and Hohmann, 1990; Sternberg the Cretaceous Housigou formation. The resistivities of the shallow
et al., 1988; Tournerie et al., 2004, 2007). Generally, transient electro- layers from the well logging were used as constraints during the inver-
magnetic (TEM) measurements are utilized to correct such effects. Set- sion. Thus a reliable representation of the resistivities can be obtained
ting a large error for the apparent resistivity during inversion is also an from the inversion.
effective method to accommodate the static shift (Ogawa, 2002). We There are many reviews regarding selecting different polarization
had no access to any TEM data during the survey, so we tried a correc- data for 2-D inversions (Berdichevsky et al., 1998; Jones, 2006; Ledo
tion based on the principle of similarity between the MT curves of adja- et al., 2002; Tauber et al., 2003; Unsworth et al., 2004; Wannamaker
cent sites (Wang, 1992). To be specic, we rst calculated the average et al., 1984). By changing the mesh size of the initial models with differ-
value of the apparent resistivity data in the high-frequency band, for ent polarization data, we compared the obtained model roughness and
Site B5 and its adjacent sites without the static shift. Then, we moved the size, shape, and boundaries of the electrical features resolved by the
the apparent resistivity curves of Site B5 as a whole downwards or up- data to determine which polarization mode data can delineate stable
wards according to the difference between the average values. models. Finally, we chose the joint TE + TM-mode inversion as the re-
sult which generates the smoothest resistivity model consistent with
4. MT inversion the data, and all structures in the resulting model are required by the
data. The proles were extended on left and right sides during the
The above-mentioned analysis demonstrates that the subsurface modeling. The extended parts were intercepted after the inversions.
electrical structure of the study area can be approximated by 2-D char- Fig. 12 shows the nal models together with the site-by-site RMS
acterization. In view of some successful cases of geothermal eld explo- values. The overall RMS of the three proles is approximately 2.2, and
ration (Bai et al., 2001; Harinarayana et al., 2006), 2-D inversions are the RMS of an individual site is always below 2.5. The RMS values
available. Two-dimensional inversion schemes can be applied if the show good correspondence between the raw data and the model
MT data validate a two-dimensional structure along some or all of the responses.
prole (Harinarayana et al., 2006). Conventional 3-D trial-and-error for- To determine the sensitivity of the data, a 2D synthetic model
ward modeling of limited soundings may not always lead to a less un- (Fig. 13a) was designed and its model responses in the range of 8000
equivocal model in the absence of accurately quantied a priori 0.01 Hz were calculated with the forward modeling, and then the
information about the study area (Bai et al., 2001). model was recovered from the synthetic data with the NLCG inversion
We employed the nonlinear conjugate-gradient algorithm code of using the TE + TM polarization mode. The result (Fig. 13b) shows that
Rodi and Mackie (2001) to carry out the 2-D inversions. The inversion the inversion can recover the narrow-topped structure. The densely dis-
code has been previously tested and proven to exhibit stable conver- tributed contours on both sides of the structure indicate the boundaries
gence in practical applications (Xiao et al., 2011, 2013). This algorithm separating high resistivity and low resistivity.
L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765 57

315 45

270 90
0 10 20 30

225 135

180

(a)

315 45

270 90
0 5 10 15 20 25

225 135

180

(b)

315 45

270 90
0 5 10 15 20 25

Fig. 8. Induction arrows in unrotated geomagnetic coordinates for the stations (redreal
induction arrows; blueimaginary induction arrows). (a) Prole Line-A. (b) Prole Line- 225 135
B. (c) Prole Line-C.

180

2D forward modeling was then performed to test whether the resis- (c)
tivity and shape beneath Site C2 are workable. The forward model was
constructed by replacing the area with a rectangle. Such a way has Fig. 9. Rose diagrams depicting the electric strike directions for all frequency bands.
been proved effective to help determine the inversion (Xiao et al., (a) Prole Line-A. (b) Prole Line-B. (c) Prole Line-C.
2012). The 1000 m rectangle is at the depth of 250 m and with the
bottom depth of 4000 m and width of 450 m. The responses of the
model were calculated so as to get the RMS mist of each site on Line-
C (Table 2). The RMS values of the model beneath Sites C2 and C3 are
58 L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765

0 0

315 45 315 45

270 90 270 90
0 1 2 3 4 5 0 2 4 6

225 135 225 135

180 180

(a) (b)
0 0

315 45 315 45

270 90 270 90
0 1 2 3 4 5 0 2 4 6

225 135 225 135

180 180

(c) (d)

Fig. 10. Rose diagrams depicting the electric strike directions averaged for 4 bands (Prole Line-C). (a) 30010 Hz. (b) 101 Hz. (c) 10.1 Hz. (d) 0.10.01 Hz.

larger than those of the inversion result in Fig. 12c. The superior ts in top depth of the structure is approximately 250 m and the bottom
the inversion indicate that the relatively-low, narrow-topped structure depth is at least 3500 m. The top of the structure is beneath Site C2.
is likely the more reliable representation of the subsurface. The high-resistivity parts may correspond to the granites or granodio-
The effective detection depth of the MT method is related to the rites from the basement. Beneath Site C2, an intermediate resistivity
penetration depth. The penetration depth is approximately 47.959 km layer with a thickness of approximately 100 m exists in between the
and the equivalent depth of investigation is approximately 33.943 km structure and conductive cap layer. The western part of the basement
when the frequency is given 0.011 Hz (the minimum frequency) and re- ascends eastwards, with the top's burial depth changing from 700 m
sistivity is 100 m. The penetration depth is approximately 15.166 km to 200 m. The eastern part thickens eastwards and north-eastwards.
and the investigation depth is approximately 10.734 km when the resis- The geoelectrical distribution of the structure can be well delineated
tivity is changed to 10 m. Therefore, the resistivity models can be with lengthwise sections along the survey proles in Fig. 14. The 2-D re-
well resolved down to 4 km even though the longest prole is only sistivity models in Fig. 14 are also obtained with the NLCG inversion
2.5 km long. code using joint TE + TM impedance data. The models also show a dis-
The surface and shallow layers (lower than 500 m in depth) of the tinct structure whose shape and resistivity features may indicate a pos-
three proles have the lowest resistivity (approximately 440 m). sibly connected conductive zone in this area.
The low-resistivity layer of Prole Line-A is the thickest, and that of The electrical slices at different depths are shown in Fig. 15. The ren-
Prole Line-C is the thinnest, from 50 m to 200 m in depth. The low- dering views of the proles (Fig. 16) clearly illustrate the perspective
resistivity formations correspond to the Quaternary clay, Neogene imaging of the low-resistivity structure and look intuitively direct
glutenite, and Cretaceous Hailang Group formations which have a when compared with common electrical slices. The images were creat-
maximum thickness of approximately 800 m. Beneath Site C2, the ed using 2-D Gaussian transfer functions and volume rendering (Zhou
conductive cap layer has a thickness of approximately 120 m. The layers and Hansen, 2013). The rendering display is a supplementary represen-
with denser resistivity contours (resistivity of approximately 40 tation of the slices as we can take a glance at the structure from different
1000 m) are underneath the lowest-resistivity layers. In Fig. 12a, b, perspectives. To highlight the structure, Fig. 17 only depicts the render-
and c, deeper formations (deeper than 500 m) are characteristic of rel- ing characterization of the anomalies with resistivities in the range 126
atively low resistivity in between the highly resistive parts (resistivity of 501 m (i.e., 2.12.7 in Lg values). It can be seen that at the depth of
approximately 160010,000 m). A relatively low-resistivity structure 3500 m the base of the structure has an area of approximately
(resistivity of approximately 100600 m) in Fig. 12c is cone-shaped 2,250,000 m2. The bottom depth of the reservoir can be even deeper ac-
with a narrow top and wide bottom. The upper width of the structure cording to Figs. 16 and 17. The western part of the structure's base is
is approximately 350 m and the lower is approximately 800 m. The thicker than the eastern part.
L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765 59

under Site C2 that is embedded in the highly resistive basement and


overlain by the conductive sedimentary cover. What is the origin of
the relatively low resistivities? Three causes are estimated as below.

(1) Fractured zones may be present in the basement. Usually, the re-
sistivity of unaltered and non-fractured granites or granodiorites
is approximately 5 103106 m. When they are fractured or
lled with uids, especially hot water, their resistivity may de-
crease to a few hundred or even lower than a hundred m.
Granites and granodiorites have very low porosity and perme-
ability and thus have very poor water abundance. Therefore,
only if structural ssures or fractures are present, such rocks
can transmit water and heat. According to the exploration results
of the controlled-source audio-frequency electromagnetics and
MT (Ye, 2011), the fractured granite basement at the Jingbo
lake graben, Ning'an, has relatively low resistivity in the range
1001000 m, as the basement is fractured and water-
bearing. The well logging data of the Jingbo lake graben show
that the resistivity of the late Variscan-age granite geothermal
reservoir is 133.8506.7 m. The anomalous structure in Hailin
has the resistivity of 100600 m. Thus, it is possible that the
basement under Site C2 has ssures or fractures and contains
hot water, producing the relatively low resistivities observed.
The MT sites are very close to the Hailang River Fault. The fault
may result in the fractured zones.
(2) The rendering views (Figs. 1617) indicate that the structure
looks like a shoot. Did magma in the deep intrude into the base-
ment in the form of a shoot? As mentioned in a previous section,
under the northwestward subduction stress eld of the Pacic
Plate, magma intruded in different directions and periods during
the magmatic activities. In this survey area, magma moved along
the Hailang River Fault and condensed after intrusion. The intru-
sive rocks in the area are mainly granites or granodiorites. Even
though the rocks are hot, they cannot produce the relatively
low resistivities. When the rocks are fractured and lled with
hot water, they may accord with the MT result.
(3) The Cretaceous water-bearing formation may persistently sub-
side and form a stable reservoir because of the continuous and
strong structural activities of the faults since the Mesozoic.
Controlled-source audio-frequency electromagnetic proles in-
dicate that Cretaceous sandstones subsided and formed a large-
scale geothermal reservoir in the Huinan section along the
DunhuaMishan Fault due to the strong fault structural activities
(Ding et al., 2014). The Huinan section is 350 km away from the
Hailin area. The Cretaceous water-bearing formation in the Hailin
survey area consists mainly of sandstone, glutenite, and siltstone,
interbedded with thin layers of mudstone. Usually, sandstones or
glutenites have resistivities of 101000 m. When they are
water- and mud-bearing, their resistivities can be much lower.
In combination with the resistivities estimated by Archie's law,
it is not probable that such a subsiding Cretaceous water-
bearing formation can produce the relatively low resistivities ob-
served. In addition, according to the geological setting, the bot-
tom depth of the structure indicates it is not a Cretaceous
formation. The inversion and rendering displays (Figs. 1217)
show that the structure may connect with a deeper low-
resistivity source at the depth larger than 3500 m.
Fig. 11. Map of the MT sites with arrows showing mean strike direction in four frequency
bands. (a) 30010 Hz. (b) 101 Hz. (c) 10.1 Hz. (d) 0.10.01 Hz.

It is most probable that a combination of the fault activity and


5. Discussion magma intrusion formed the basement fractures where the water
from the Hailang River and atmospheric precipitation seeped in and
5.1. Geophysical interpretation was heated. This is the cause of the relatively low resistivity.
The electrical conductivity features revealed by the magnetotelluric
The 2-D inversions (Figs. 12, 14) and rendering views (Figs. 1617) data indicate favorable storage space for a zonal geothermal reservoir.
reveal the presence of a relatively low-resistivity, anomalous structure The size and shape of the anomalous structure appear to agree with a
60 L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765

geothermal source's requirements and are similar to those of the geo- section below the clay cap and highly resistive surroundings are some-
thermal system depicted by Wright et al. (1985). According to Wright how indicative for a high-temperature geothermal system (Johnston
et al. (1985), a low-resistivity zone produced by the brines and clays et al., 1992) and represent a possible drilling target. Although the lithol-
capping a geothermal system provides a feature that can be easily de- ogy and resistivity of the capping formation and the structure below
tectable by electromagnetic methods; the relatively higher resistivity Hailin are not completely consistent with those of the system described

RMS
2
1
0

West A1 A2 A3 A4 A5 East
0

-500

-1000
ohm-m
Lg
-1500 3.8
Depth(m)

3.4
-2000
3

2.6
-2500
2.2

-3000 1.8

1.4
-3500
1

-4000 0.6
0 500 1000 1500
Distance(m)

(a)

3
RMS

2
1
0

West B1 B2 B3 B4 B5 B6 East
0

-500

-1000
ohm-m
Lg
Depth(m)

-1500 3.8

3.4
-2000
3

2.6
-2500
2.2

-3000 1.8

1.4
-3500
1

-4000 0.6
0 500 1000 1500 2000 2500
Distance(m)

(b)
Fig. 12. 2-D inversion of the proles (joint TE + TM mode). (The resistivity values are the results after the common logarithm calculation. The site-by-site RMS values are shown above the
models.) (a) Line-A. (b) Line-B. (c) Line-C.
L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765 61

RMS
2
1
0

West Drilling well East


C1 C2 C3 C4 C5
0

-500

-1000
ohm-m

-1500 3.8
Depth(m)

3.4
-2000
3

2.6
-2500
2.2

-3000 1.8

1.4
-3500
1

-4000 0.6
0 500 1000 1500 2000
Distance(m)

(c)
Fig. 12 (continued).

by Wright et al. (1985) and Johnston et al. (1992), the shape (narrow
top and wide bottom) and size of the structures are alike, indicating fa-
vorable storage space for geothermal water.
Nevertheless, the MT result only shows the electrical resistivity fea-
ture and structure. It has to be combined with geological conditions to
predict the occurrence of a geothermal reservoir.

5.2. Geological factors of geothermal sources

The generation of geothermal resources needs not only geothermal


gradient variations but also suitable structures, magmatic (volcanic) ac-
tivity, storage space for groundwater, migration pathways, and closure
(capping formation) conditions. According to the geological background
introduced in a previous section, these conditions are present in this
study area.
The fractured zones of the faults in Mudanjiang facilitate the migra-
tion and supplementation of heat and groundwater (Sui et al., 2011). As
mentioned before, the sedimentary formations with large thicknesses Fig. 13. A synthetic model (a) and the inversion (joint TE + TM mode) (b). The model is
prevent subsurface heat from diffusing; rocks with some porosity are similar to the structure of Line-C (Fig. 12c) and its sides and bottom are largely extended
when modeling. The model is of 100 m homogeneous half-space in the inversion. After
helpful to divert water. Poor thermal conductivity enables the cap
the inversion only the part is shown in (b). The RMS is 1.1.
rocks to insulate the heat. The Hailang River that ows through the
MT survey area provides sufcient groundwater supplementation. The
river water and atmospheric precipitation permeate the ground water temperature may have reached 120 C at a depth of 2000 m.
through the surface faults and fractured zones and slowly travel into The hydraulic discharge of the well is 100 m3/h.
the reservoir. The water continuously absorbs the heat conveyed from The correspondence between the MT and drilling results (tempera-
the faults and the granite basement. ture and depth) is very meaningful. Taking into account the well's tem-
An exploratory well (see the dark blue arrow in Fig. 12c) was drilled perature records when interpreting the posterior resistivity distribution
near Site C2 after the MT survey, reaching almost 655 m depth. This al- can be useful to reduce the uncertainty of the EM inversion results
lows the resistivity distribution of the MT proles to be correlated with
the drilling and lithologic logs obtained from the well (the logs are not Table 2
permitted to be published, so they are not shown in this paper). The RMS ts.
drilling result of the exploratory well revealed that hot water began to
Site C1 Site C2 Site C3 Site C4 Site C5
be injected where the drilling depth was 172 m; the groundwater tem-
perature was 42 C when the drilling depth was 360 m and as high as Inversion model in Fig. 12c 1.10 1.08 0.93 1.18 1.24
Forward model 1.41 3.35 2.85 1.37 1.29
85 C when the depth was 655 m. The drilling logs indicated that
62 L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765

C1 C2 C3 C4 C5

B1 B2 B3 B6
B4 B5

A1 A2 A3 A4
A5

A1 B1 C1 A2 B2 C2 A3 B3 C3 A4 B4 C4 A5 B5 C5
0 0 0 0 0

-500 -500 -500 -500 -500 ohm-m


Lg
-1000 -1000 3.8
-1000 -1000 -1000
3.4
-1500 -1500 -1500
Depth(m)

-1500 -1500
3

-2000 -2000 -2000 -2000 -2000


2.6

-2500 -2500 -2500 2.2


-2500 -2500
1.8
-3000 -3000 -3000 -3000 -3000
1.4

-3500 -3500 -3500 -3500 -3500


1

-4000 -4000 -4000 -4000 -4000 0.6


0 500 1000 1500 2000 0 500 1000 1500 2000 0 500 1000 1500 2000 0 500 1000 1500 2000 0 500 1000 1500 2000
Distance(m)

Fig. 14. 2-D inversion of the lengthwise survey lines (joint TE + TM mode). (The resistivity values are the results after the common logarithm calculations).

(especially when using 1-D or 2-D inversion schemes) (Spichak and (2) Heat from crust and mantle sources has been delivered through
Manzella, 2009). By comparing the resistivity distributions with the rocks or faults to the surface, providing stable deep-seated ther-
temperature, the validity of the location and dimensions of the estimat- mal sources for the regional geothermal eld. The burial depth
ed reservoirs could be conrmed (Asaue et al., 2006). of the Moho in the study area is approximately 3031 km
Based on the geological conditions, we infer four geological factors (Jiang et al., 2006). The comparatively thin crust facilitates the
that inuence the geothermal sources of Hailin. overow of heat outwards.
(3) In view of the heat generation in the Ning'an reservoir (Sui et al.,
(1) Large areas of granites form the major thermal source in the 2011), we deduce that the faults and their subsidiary faults in
study area. Different types of rocks have different amounts of ra- Hailin may generate heat due to mechanical friction during
dioactive elements; among them, granites have the highest heat structural activity, thus forming another heat source for the geo-
production rate. The radioactive heat produced by the granites thermal site. In addition, the faults are favorable channels for the
has strong sustainability and thus has affected the geothermal movement of deep heat into the upper strata.
eld in different development stages. Granites are also important (4) As mentioned before, the content ratio of radioactive elements in
heat sources in other areas of China. For example, the geothermal the Ning'an Basin is relatively high, which helps the formation of
reservoir of the Rehai (hot sea) eld in Tengchong, Yunan Prov- heat resources.
ince, is made of the Late Cretaceous granite (Liao et al., 1991).
The geotemperature eld of the Songliao Basin is mainly deter-
mined by large areas of granite in the basement (He, 2012).

Fig. 15. Slice plot of the proles. Fig. 16. Rendering display of the proles.
L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765 63

Acknowledgments

We thank Dr. Bernhard Friedrichs and Dr. Ulrich Matzander of


Metronix Company, Germany, for their earnest help and valuable sug-
gestions. We acknowledge the reviewers and editors for their construc-
tive remarks and suggestions. Prof. Randy Mackie is acknowledged for
providing the free version of the NLCG inversion code. We thank Prof.
Kong Xiangru and Prof. Chen Xiaobin for their instructive help. Dr. Xu
Ya participated in the data collection. Dr. Weijian Hu helped to modify
the gures. This work was nancially supported by the National Natural
Science Foundation of China (NSFC) (grant no. 41174091), the open
project offered by the State Key Laboratory of Petroleum Resource and
Prospecting, China University of Petroleum (PRP/open-1110), and the
National Major Project of China (2011ZX05008-006-30).

Appendix A

The PM model proposed by Perona and Malik (1990) is a thermal


diffusion partial differential equation and is applied to two-
Fig. 17. Rendering display of structures with resistivity ranging from 126 to 501 m
dimensional images:
(i.e., 2.12.75 in Lg values).
8
< ux; y; t
divg kuku ; A  1
: t
ut 0 t 0

where u(x, y, t) is the image matrix P(x, y) with respect to the space
6. Conclusion variables, x and y are the width and height of the matrix, respectively,
div is the divergence operator, is the gradient operator, u0(x) is initial
Geothermal resources are closely related to the conditions of deep signal, and t is the scale-space parameter. g(u) is the diffusion
structures. The MT method, which is one appropriate method for solv- coefcient of the diffusion equation, which is a decreasing function. In
ing deep geologic problems, has unique advantages. The reconstructed this way, the diffusion process is inhibited around the edges where
2-D images and rendering visualization meaningfully depict the geo- the gradient is high. g(u) has the form of a Lorentzian distribution
thermal structure in the area. The fault activities and magma intrusion function:
may result in the fractures of the basement, which are lled with hot
water and thus produce the relatively low resistivity (approximately
1
100600 m). The good agreement between the MT inversions and g u ; A  2
the drilling result shows that the MT method can provide useful depic- 1 kuk=k2
tions of the geothermal reservoir.
Moreover, modern 3-D modeling and inversion techniques are rec- or
ommended to unravel the deep resistivity structures and boundaries
in deep parts of the geothermal reservoir. The further work will be to in- kuk=k2
g u e : A  3
vestigate whether the 3-D modeling and imaging can avoid artifacts in-
herent in the 2-D data analysis. Because 3-D MT inversion and modeling
requires signicant computational resources and time (Newman and The scale spaces generated by these two functions are different:
Alumbaugh, 2000; Newman et al., 2003), it was more efcient to build the rst favors wide regions over smaller ones, while the second fa-
an initial 3-D resistivity model from the 2-D imaged sections. This vors high-contrast edges over low-contrast ones. The parameter k
starting model will be rened subsequently through 3-D inversion pro- is the diffusion threshold, which controls the rate that the diffusion
cess. The rendering display of the models resolved from 3-D inversions coefcient decreases as the absolute value of the image gradient in-
will be a meaningful 3-D visualization. creases. The value of k is set in relation to the gradient strength of
The modied anisotropic-diffusion-based method has been tested the edges that are to be preserved during the diffusion. When u
with the synthetic signals and MT data from Hailin. The results demon- is much larger than k, the diffusion is suppressed and the edges can
strate that the method can effectively suppress noise and perform better be preserved.
than the median ltering and conventional PM model. Nevertheless, The denoising principle of the PM model is as follows: detect the
further work is needed to address different types of real noise that are gradient variation caused by noise using the gradient operator, then
acquired in eld and adapt the suppression accordingly. With the suppress the minor gradient variation caused by noise and preserve
rapid development of industrialization and population in China, electro- the major gradient variation caused by edges using the diffusion
magnetic signals have been polluted by various interferences, and equation. More details can be seen in the literature written by
denoising will be an indispensable part of MT data processing. Noise Perona and Malik (1990) and other relevant papers, which are wide-
suppression will be very promising to help restore real signals. ly published.
The MT exploration of the geothermal reservoir can provide valuable
constraints on the physical conditions within the Mudanjiang Fault Appendix B
zone. It has guiding signicance for the exploration of other geothermal
resources along the Mudanjiang Fault. Favorable structures and path- The modied method differs from the PM model in two aspects: the
ways for water and heat can be paid attention to especially in granite diffusion equation is made for 1-D time series, and the diffusion coef-
areas. cient is adaptively selected based on the characteristics of the data.
64 L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765

The method uses the following diffusion equation: Fontes, S.L., Harinarayana, T., Dawes, G.J.K., Hutton, V.R.S., 1988. Processing of noisy
magnetotelluric data using digital lters and additional data selection criteria. Phys.
8 Earth Planet. Inter. 52, 3040.
< ux; t Gamble, T.D., Goubau, W.M., Clarke, J., 1979. Magnetotellurics with a remote magnetic
divg kuku : B  1 reference. Geophysics 44, 5368.
: t Groom, R.W., Bahr, K., 1992. Corrections for near surface effects: decomposition of the
ut 0 t 0 magnetotelluric impedance tensor and scaling corrections for regional resistivities:
a tutorial. Surv. Geophys. 13, 341379.
Groom, R.W., Bailey, R.C., 1989. Decomposition of magnetotelluric impedance tensors in
The diffusion coefcient g(u) is modied and adaptively the presence of local three-dimensional galvanic distortion. J. Geophys. Res. 94,
selected: 19131925.
Harinarayana, T., Abdul Azeez, K.K., Murthy, D.N., Veeraswamy, K., Eknath Rao, S.P.,
8 Manoj, C., Naganjaneyulu, K., 2006. Exploration of geothermal structure in Puga geo-
gkuk e
> kMux;t k=k
2
>
< ; if kuk kk thermal eld, Ladakh Himalayas, India by magnetotelluric studies. J. Appl. Geophys.
g kuk  e
kMux;t sk=k 58, 280295.
2 2
kuk=k
> e ; if kk kuk k k
>
: kuk=k
2 He, X.P., 2012. The distribution characteristics and resource evaluation of geothermal res-
g kuk e ; if kuk k k ervoir in Kedong Town. (Master dissertation), Jilin University (In Chinese).
B  2 Jiang, W.W., Zhou, L.H., Xiao, D.Q., Gao, J.R., Yuan, S.Q., Tu, G.H., Zhu, D.Y., 2006. The char-
acteristics of crust structure and the gravity and magnetic elds in northeast region of
China. Prog. Geophys. 21, 730738 (In Chinese).
where u(x, t) is the time series P(x) varying with the space variable t and Jing, J.A., Wang, Y., Li, W., 2008. The structural evolution and prognosis of uranium explo-
ration in Ning'an basin. Uranium Geol. 24, 143149 (In Chinese).
M is the median ltering. k is related to the degree of noise and is given Johnston, J.M., Pellerin, L., Hohmann, G.W., 1992. Evaluation of electromagnetic
a relatively small value (5), empiric or experimental. Provided that k methods for geothermal reservoir detection. Geotherm. Resour. Counc. Trans.
and k are given, and are obtained through curve tting (Wang 16, 241245.
Jones, A.G., 1988. Static shift of magnetotelluric data and its removal in sedimentary basin
et al., 2006). environment. Geophysics 53, 967978.
A time series is a sequence of observations that is typically measured Jones, A.G., 2006. Electromagnetic interrogation of the anisotropic earth: looking
at successive points in time and spaced at uniform time intervals. The into the earth with polarized spectacles. Phys. Earth Planet. Inter. 158,
281291.
series value is plotted on the vertical axis and the time on the horizontal Jones, A.G., Groom, R.W., 1993. Strike-angle determination from the magnetotelluric im-
axis. pedance tensor in the presence of noise and local distortion: rotate at your peril!
The gradient operator in the diffusion equations performs as an edge Geophys. J. Int. 113, 524534.
Kong, X.R., Liu, S.J., Zhang, J.J., 1991. Magnetotelluric sounding studies in the eastern area
detector. The edge detector of an image matrix is the difference calcula-
of Fujian Province. Acta Geophys. Sin. 34, 724735 (In Chinese).
tion of one matrix element with its neighbors in four directions. Howev- Ledo, J., Queralt, P., Marti, A., Jones, A.G., 2002. Two-dimensional interpretation of three-
er, the detector only considers two directions, i.e., left or right, for the dimensional magnetotelluric data: an example of limitations and resolution.
one-dimensional problem. Geophys. J. Int. 150, 127139.
Li, W., Chen, C.H., Cui, J.H., 2007. Analysis on hydrogeochemical conditions of uranium
The data values of the images are in the range from 0 to 255. Thus, metallogenesis in Hailang hollow of Ning'an basin. Uranium Geol. 23, 156163 (In
the diffusion threshold k is usually given an empiric value. However, a Chinese).
1-D seeking method is used for the 1-D time series data to select an op- Liao, Z.J., Shen, M.Z., Guo, G.Y., 1991. Characteristics of the geothermal reservoir in the
Rehai (hot sea) eld in Tengchong county, Yunan Province. Acta Geol. Sin. 1, 7385
timum k value. (In Chinese).
The diffusion equation of the modied method is equal to the 1-D McNeice, G.W., Jones, A.G., 2001. Multisite, multifrequency tensor decomposition of
form of the PM model; therefore, the modied method can inherit the magnetotelluric data. Geophysics 66, 158173.
Meju, M.A., 1996. Joint inversion of TEM and distorted MT soundings: some effective
advantages of the PM model. Nevertheless, the modication of the diffu- practical considerations. Geophysics 61, 5665.
sion coefcient makes a difference. The PM model uses only one form of Newman, G.A., Gasperikova, E., Hoversten, G.M., Wannamaker, P.E., 2008. Three-
the diffusion coefcient (Eqs. (A-2) or (A-3)) for the whole data, but the dimensional magnetotelluric characterization of the Coso geothermal eld.
Geothermics 37, 369399.
modied method uses three forms (Eq. (B-2)) according to the gradient Newman, G.A., Alumbaugh, D.L., 2000. Three-Dimensional Magnetotelluric Inversion
variation. Through the adaptive selection of the diffusion coefcient, a using Non-Linear Conjugate Gradients. Geophys. J. Int. 140 (2), 410424.
larger degree of diffusion is applied to segments with smaller gradients, Newman, G.A., Recher, S., Tezkan, B., Neubauer, F.M., 2003. Case history 3D inversion of a
scalar radio magnetotelluric eld data set. Geophysics 68, 791802.
while the PM model is applied to those with larger gradients. When the
Ogawa, Y., 2002. On two-dimensional modeling of magnetotelluric eld data. Surv.
gradient is very small, the median lter replaces the gradient operator Geophys. 23, 251273.
for the diffusion coefcient (see Eq. (B-2)). This adaptive selection can Ogawa, Y., Uchida, T., 1996. A two-dimensional magnetotelluric inversion assuming
increase the discrimination ability. Gaussian static shift. Geophys. J. Int. 126, 6976.
Padilha, A.L., Vitorello, ., Pdua, M.B., Bologna, M.S., 2006. Lithospheric and
sublithospheric anisotropy beneath central-southeastern Brazil constrained by long
References period magnetotelluric data. Phys. Earth Planet. Inter. 158, 190209.
Pellerin, L., Hohmann, W.G., 1990. Transient electromagnetic inversion: a remedy for
Alvarez, L., Lions, P.L., Morel, J., 1992. Image selective smoothing and edge detection by magnetotelluric static shifts. Geophysics 55, 12421250.
nonlinear diffusion. SIAM J. Numer. Anal. 29, 845866. Pellerin, L., Johnston, M.J., Hohmann, W.G., 1996. A numerical evaluation of electromag-
Asaue, H., Koike, K., Yoshinaga, T., Takakura, S., 2006. Magnetotelluric resistivity modeling netic methods in geothermal exploration. Geophysics 61, 121130.
for 3D characterization of geothermal reservoirs in the Western side of Mt. Aso, SW Perona, P., Malik, J., 1990. Scale space and edge detection using anisotropic diffusion. IEEE
Japan. J. Appl. Geophys. 58 (4), 296312. Trans. Pattern Anal. Mach. Intell. 12, 629639.
Bahr, K., Simpson, F., 2005. Practical Magnetotellurics. Cambridge University Press. Qi, M., 2006. Three Dimensional visualization of geological structure based on geophysical
Bai, D.H., Meju, A.M., Liao, Z., 2001. Magnetotelluric images of deep crustal structure of the eld data. (PhD thesis), IGGCAS (In Chinese).
Rehai geothermal eld near Tengchong, southern China. Geophys. J. Int. 147, Rodi, W., Mackie, R.L., 2001. Nonlinear conjugate gradients algorithm for 2-D
677687. magnetotelluric inversions. Geophysics 66, 174187.
Berdichevsky, M.N., Dmitriev, V.I., 2008. Models and Methods of Magnetotellurics. Schmucker, U., 1970. Anomalies of geomagnetic variations in the southwestern United
Springer. States. Bull. Scripps Inst. Oceanogr. 13, 165.
Berdichevsky, M.N., Dmitriev, V.I., Pozdnjakova, E.E., 1998. On two-dimensional interpre- Simpson, F., Bahr, K., 2005. Practical Magnetotellurics. Cambridge University Press,
tation of magnetotelluric soundings. Geophys. J. Int. 133, 585606. Cambridge.
Chen, L.S., Wang, G., 1990. Magnetotellurics. Geological Publishing House, Beijing. Smith, S.R., O'Connell, D.M., 2005. Interpolation and gridding of aliased geophysical data
Chen, X.B., Zhao, G.Z., Zhan, Y., Tang, J., 2004. Analysis of tipper visual vectors and its ap- using constrained anisotropic diffusion to enhance trends. Geophysics 70, V121V127.
plication. Earth Sci. Front. 11, 626636 (In Chinese). Spichak, V., Manzella, A., 2009. Electromagnetic sounding of geothermal zones. J. Appl.
Corwin, F.R., Hoover, B.D., 1979. The self-potential method in geothermal exploration. Geophys. 68, 459478.
Geophysics 44, 226245. Spichak, V., Zakharova, O., Rybin, A., 2007. Estimation of the sub-surface temperature by
Ding, Z.Q., Xu, Z.H., Su, Y.F., Guan, S.Y., Chen, Q., 2014. The possibility of structure and oc- means of magnetotelluric sounding. XXXII Workshop on Geothermal Reservoir Engi-
currence geothermal resources in DunhuaMishan fault zone (Huinan section). Jilin neering, Stanford, USA (Expanded Abstracts).
Geol. 33, 98102 (In Chinese). Sternberg, K.B., Washburne, C.J., Pellerin, L., 1988. Correction for the static shift in
Fehmers, G.C., Hocker, C.F.W., 2003. Fast structural interpretation with structure-oriented magnetotellurics using transient electromagnetic soundings. Geophysics 53,
ltering. Geophysics 68, 12861293. 14591468.
L. Zhang et al. / Journal of Applied Geophysics 118 (2015) 4765 65

Sui, X.W., Zhang, J., Shi, F.J., Zhao, J.C., 2011. Evaluation on the geothermal resources in Wang, Z.Y., Huang, C.X., 2014. The geothermal resource evaluation of the Jingbohu Basin
Jingbo Lake Graben of the DunMi Fault Zone in Heilongjiang Province. Geoscience in Mudanjiang, Heilongjiang Province. Proceedings of the 2014 National Science and
25, 377383 (In Chinese). Technology Work Conference, pp. 208210 (In Chinese).
Swift, C.M., 1967. A magnetotelluric investigation of electrical conductivity anomaly in the Wang, Y., Zhang, L.P., Li, P.X., 2006. An improved algorithm of anisotropic diffusion
southwestern United States. (PhD thesis). Massachusetts Institute of Technology, smoothing lter. J. Image Graph. 11, 210216 (In Chinese).
Cambridge, MA. Wannamaker, P.E., Hohmann, G.W., Ward, S.H., 1984. Magnetotelluric responses of three-
Tauber, S., Banks, R., Ritter, O., Weckmann, U., 2003. A high-resolution magnetotelluric dimensional bodies in layered earths. Geophysics 49 (9), 15171533.
survey of the Iapetus Suture zone in southwest Scotland. Geophys. J. Int. 153, Wright, P.M., Ward, S.H., Ross, H.P., West, R.C., 1985. State-of-the-art geophysical explora-
548568. tion for geothermal resources. Geophysics 50, 26662696.
Telford, W.M., Geldart, L.P., Sheriff, R.E., 1990. Applied Geophysics. Second edition. Xiao, Q.B., Cai, X., Liang, G., Xu, X., Zhang, B., 2011. Application of 2D magnetotelluric
Cambridge University Press, Cambridge. methods in a geological complex area, Xinjiang, China. J. Appl. Geophys. 75, 1930.
Tournerie, B., Chouteau, M., Marcotte, D., 2004. Estimation and removal of the MT static Xiao, Q.B., Zhang, J., Wang, J.J., Zhao, G.Z., Tang, J., 2012. Electrical resistivity structures be-
shift effect using geostatistical methods (Workshop on Electromagnetic Induction tween the Northern Qilian Mountains and Beishan Block, NW China, and tectonic im-
in the). plications. Phys. Earth Planet. Inter. 200201, 92104.
Tournerie, B., Chouteau, M., Marcotte, D., 2007. Magnetotelluric static shift: estimation Xiao, Q.B., Zhang, J., Zhao, G., Wang, J., 2013. Electrical resistivity structures northeast of
and removal using the cokriging method. Geophysics 72, F25F34. the Eastern Kunlun Fault in the Northeastern Tibet: tectonic implications.
Unsworth, M., Wei, W., Jones, A.G., Li, S., Bedrosian, P., Booker, J., Jin, S., Deng, M., Tan, H., Tectonophysics 601, 125138.
2004. Crustal and upper mantle structure of northern Tibet imaged with Ye, G.H., 2011. Foreground analysis of geothermal resources in Mudanjiang. Guide of Sci-
magnetotelluric data. J. Geophys. Res. 109, B02403. tech Magazine 8 p. 192 (In Chinese).
Volpi, G., Manzella, A., Fiordelisi, A., 2003. Investigation of geothermal structures by Zhang, M.S., Peng, X.D., Sun, X.M., 1998. The Paleozoic tectonic geographical pattern of
magnetotellurics (MT): an example from the Mt. Amiata area, Italy. Geothermics northeast China. Liaoning Geol. 2, 9196 (In Chinese).
32, 131145. Zhou, L., Hansen, C.D., 2013. Transfer function design based on user selected samples for
Wan, B., Zhong, Y.Z., 1997. Feature analysis and division of new tectonic movement in intuitive multivariate volume exploration. Proceedings of the 2013 IEEE Pacic Visu-
northeast China. Seismol. Res. Northeast China 13, 6475 (In Chinese). alization Symposium (PacicVis), pp. 7380.
Wang, J.Y., 1992. Problem about static correction in magnetotellurics. Geol. Sci. Technol.
Inf. 11, 6976 (In Chinese).