2.1 Geology and geodynamic activity

The Cenozoic Eger Rift with the central Eger Graben, the NNW–SSE-trending Mariánské Lázně fault zone (MLF), and the Cheb–Domažlice Graben are prominent tectonic structures of the Bohemian Massif, which is the eastern part of the European Cenozoic Rift System (Bankwitz et al., 2003b; Malkovský, 1987; Ziegler, 1992; Peterek et al., 2011). The Eger Rift contains several basins (e.g., Cheb Basin, Sokolov Basin, Most Basin) with similar sedimentary and tectonic evolution (Pešek et al., 2014). The investigation area, the geodynamically active Cheb Basin, a shallow Neogene intra-continental basin with a maximal depth of approximately 350 m, was formed at the intersection of the NE–SW-striking Eger Graben and the NNW-striking Cheb–Domažlice Graben (Špičáková et al., 2000; Peterek et al., 2011). The Cheb Basin is bounded on its eastern side by the morphologically distinct scarp of the NNW–SSE-trending Mariánské Lázně fault, the down-dipping Smrčiny–Fichtelgebirge Mountains to the west, and the Bohemian Forest to the south (Fig. 1, Peterek et al., 2011; Bussert et al., 2017). At the west and east border of the Cheb Basin, the basement has an offset of more than 200–400 m. To the north and south, the bottom of the basin thins out gradually to the surface (Bankwitz et al., 2003b; Rojik et al., 2014).

Babuška et al. (2007) point out that the Cheb Basin is located above a triple junction of the Variscan crustal units of the Saxothuringian in the northwest, the Teplá–Barrandian in the central region, and the Moldanubian in the southeast. The basin is embedded into Proterozoic and Paleozoic magmatic and metamorphic rocks of the northwestern Bohemian Massif – predominantly granites, gneisses, mica schists, and phyllites. The sedimentary fill of the Cheb Basin around the area of interest itself consists mainly of less than 300 m of continental clastics (representing debris of these rocks; Bussert et al., 2017; Fig. 1) and overlies the deeply weathered mica schists with interbeds of metaquartzite, metabasite, and crystalline limestone that are intruded by granitoid plutons (Variscan Smrčiny, Fichtel, and Žandov plutons; Pešek et al., 2014). Several uplift and subsidence events due to varying extensional and compactional stress within the Eger Rift since the Eocene affected the sedimentation within the basin (Peterek et al., 2011; Pešek et al., 2014; Rojik et al., 2014; Bussert et al., 2017). After the local deposition of clays and sands in the Eocene (Staré Sedlo formation), sedimentation continued with the deposition of Oligocene to Miocene gravel, sand, and clays (called the Lower Argillaceous–Sandy formation or the Lower Clay–Sand formation). During the Lower Miocene, wetlands dominated the area and led to the deposition of the coal- and lignite-bearing Main Seam formation. As the result of ongoing tectonic activity, a lake developed in which the clay-dominated Cypris formation was deposited. After a hiatus, sedimentation started again in the Pliocene with lacustrine clays, sands, and gravels of the Vildštejn formation and continued without an obvious break into the Quaternary.

Currently, the area around Nový Kostel (Fig. 1) is the most active earthquake swarm zone in western Bohemia–Vogtland (Fischer et al., 2014). The activity at the Nový Kostel focal zone is supposed to be related to the reactivation of a system of faults, e.g., at the intersection between the NNW–SSE-trending MLF and the N–S-trending PPZ. The earthquake foci are located at depths between 6 and 13 km and clustered along vertical faults, forming an almost continuous, about 15 km long belt striking NNW to SSE and steeply dipping westwards (Fischer and Michálek, 2008; Fischer et al., 2014). Normal and strike-slip faulting are the typical focal mechanisms for these intra-plate events here. Most of the micro-earthquake hypocenters are aligned in a N–S direction and thus follow the course of the PPZ, whereas the NNW–SSE-striking MLF seems to be only partially seismically active (Bankwitz et al., 2003b; Fischer et al., 2014). The PPZ forms an escarpment of more than 20 m height in Pliocene–Pleistocene sediments and has probably been active since the late Pleistocene (Bankwitz et al., 2003b; Peterek et al., 2011; Bussert et al., 2017). Strike-slip faults with a vertical component run across the basin in an E–W direction (e.g., Nová Ves fault) according to Bankwitz et al. (2003a). The combination of seismological and especially hydrological analyses points out that the Nový Kostel zone is also part of the gas uplift system and must be linked to the near-surface water flux. The model, which Neunhöfer and Hemmann (2005) proposed, provides an explanation for the active ascent of fluids in relation to the phenomenon of earthquake swarms. The model takes a special two-phase system formed by water and CO₂, in contrast to other mixed models (Bräuer et al., 2008, 2009), into account. Furthermore, Horálek and Fischer (2008) assumed that ascending crustal fluids could play a key role in the alteration of the preexisting, favorably oriented faults from a subcritical to critical state due to pore pressure increase. Although ascending fluids from deep crustal root zones are considered to be the main factors inducing recurring earthquake swarms by pore pressure increase (Špičak and Hóralek, 2001; Weinlich et al., 1998; Heinicke and Koch, 2000; Weise et al., 2001; Bräuer et al., 2005, 2009; Kämpf et al., 2013; Fischer et al., 2014), the relation between earthquake swarms and the source of CO₂, CO₂ ascent, and degassing is still a matter of discussion (Babuška et al., 2016).

One of the main fluid discharge centers for carbon dioxide via mofettes at the surface is located approx. 10 km south of Nový Kostel along the course of the PPZ (Bublák and Hartoušov mofette fields). Only isolated CO₂ vents and mineral springs are found close to the MLF (e.g., Dolni Častkov mofette). The numerous cold CO₂ emanations with >99 % vol CO₂ and a mantle signature (He and N isotopes) are supposed to be generally connected to the seismic activity and to stem from upper mantle reservoirs (Weinlich et al., 1998; Geissler et al., 2005; Bräuer et al., 2009, 2011). From the high gas flux rates and high ³He∕⁴He ratios, the mofette field Bublák–Hartoušov appears to act as deep-seated fluid migration zone along the PPZ (Bräuer et al., 2011; Kämpf et al., 2013). The tectonic setting of the area has a great influence on the increased degassing of CO₂ at the surface. Since the early work of Irwin and Barnes (1980), it has become evident that a close relationship exists between the tectonic activity and anomalous crustal emissions of CO₂. Due to their hydraulic permeability, faults can act as preferential pathways for the upward migration and release of deep fluids to the atmosphere in this area (Bankwitz et al., 2003a; Geissler et al., 2005). At the surface, CO₂ emission occurs often at gas vents with diameters <1 m (Kämpf et al., 2013; Nickschick et al., 2017), with high flux rates and in moderate amounts diffusely over the larger area in general (Kämpf et al., 2013; Nickschick et al., 2015, 2017, see also Sect. 2.2). However, the deep structure, geometry, and lateral extension due to the depth of the fluid pathways in the crust layers are still unknown. Despite the geodynamic–geophysical and especially seismological research (Švancara et al., 2000; Růžek and Horálek, 2013; Fischer et al., 2014) in this area, many questions about the settings for the fluid regime and the generation of the earthquake swarms remain unanswered. Besides the local and regional stresses, as well as contrasts in rheological rock properties, the fluid movement and distribution are essential factors influencing the seismicity of the region. One peculiar phenomenon is the spatial separation of the earthquakes near Nový Kostel and the CO₂ degassing near Hartoušov, despite having a similar source behind them. However, in May 2018, a cluster of several (>70) small-magnitude earthquakes was registered (Czech PEPIN seismological catalog; http://www.ig.cas.cz, last access: 7 November 2019) a few hundreds of meters to the NE of the mofette field Hartoušov.

Figure 2Lithological transect along the large-scale profile based on the descriptions of boreholes from the Czech Geological Survey (formerly GEOFOND). The question mark indicates an area of unknown lithology and the uncertainty of whether Main Seam and Lower Clay (or Argillaceous) and/or Sandy formations are present in this area. P1–P3 mark the locations of the small-scale ERT profiles. For each drill's location, please refer to Fig. 1.

Download

2.2 Existing geophysical results and lithological data

From previous geoelectrical investigations, results from several 2-D ERT profiles with lengths of 100–750 m and an investigation depth of approx. 80–100 m across the main faults of the Cheb Basin (MLF and PPZ; Fig. 1) are available (Flechsig et al., 2008, 2010; Fischer et al., 2014; Nickschick et al., 2015, 2017; Blecha et al., 2018). The obtained resistivity models reveal the characteristics and width of the fault zones in the shallow subsurface by means of resistivity anomalies, variations in sediment thickness, and vertical layer displacement. Significant resistivity anomalies in the subsurface reveal the location of both the MLF and PPZ, and typical conductive features indicate potential fluid transport paths and regions with mineral alteration. Essentially, both fault zones are characterized by an extended subsurface region (100–250 m) controlled by multiple, more or less parallel sub-faults with different strike angles. As a local comparative geoelectric (3-D small-scale ERT), soil gas, and sediment study of a CO₂ degassing vent in the Hartoušov mofette field, near-surface structures to a depth of 20 m were investigated by Flechsig et al. (2008). The investigations reveal substantial structural features that are to be directly or indirectly related to high CO₂ flow (anomalies of electrical resistivity, self-potential, and sediment properties). With the aim of reaching deeper structures up to 5 km, several magnetotelluric investigations in the western margin of the Bohemian Massif and along the 9HR seismic profile (Cerv et al., 1997, 2001; Pícha and Hudeková, 1997; Di Mauro et al., 1999) have been carried out since 1990, resulting in very coarse conductivity models.

Recent information about the regional distribution of electrical resistivity up to 25 km of depth came from a 2-D magnetotelluric experiment on a 50 km long N–S profile with 25 stations crossing the Cheb Basin in 2017 (Muñoz et al., 2018). The most prominent deep-reaching structure is a channel of higher conductivity compared to the surroundings, which extends from the surface at the mofette field of Bublák–Hartoušov into the lower crust (approximately 25 km) to the north, possibly correlated with the hypocenters of the seismic events of the Nový Kostel focal zone. This channel has been interpreted by other authors as a pathway from a mid-crustal fluid reservoir to the surface along deep-reaching faults. Whereas the overall resistivity is very high (>500–1000 Ω m) in parts of the model, very low resistivity (<30 Ω m) could be found near the surface at the mofette fields of Bublák and Hartoušov and their feeding system. Further relevant data and information from other geophysical methods for interpretation of the measured ERT profile are not available or not in the necessary scale.

To interpret the subsurface resistivity situation around our survey target, borehole descriptions from the Czech Geological Survey (formerly GEOFOND) were gathered. In order to establish a conception of the encountered lithologic units in this experiment, we generated a 2-D transect based on the borehole data to a depth of 50 to 400 m (Figs. 1 and 2). From the available drills in the investigation area, we selected 20 that provided sufficient depth and were closest to our ERT profile. The GeODin software was used to generate the transect that can be seen in Fig. 2. Please note that none of these drills have reached the crystalline phyllite in its unweathered state and only describe the basement phyllite as weathered or highly weathered. In addition to this geological constraint, we regarded the results from Dobeš et al. (1986): their report contains valuable petrophysical information from previous studies about the different stratigraphic units in and below the Cheb Basin, which we have summarized in Table 1. The phyllitic–granitic basement is characterized by low porosities of less than 5 % compared to the sedimentary deposits on top, which feature porosities of 15 %–30 %. Resistivity, however, may vary drastically, depending on heterogeneities within the sediments and whether fluids such as mineral waters or CO₂ are present or not, and the report does not specifically state where the samples were taken from. For this area, Bussert et al. (2017) provide additional information. Not only do they mention the occurrence of highly mineralized water in the central part of the HMF, but their geophysical log of the HJB-1 drill also reveals resistivities of 5–10 Ω m for the Cypris formation and 10–20 Ω m for the topmost part of the weathered phyllites. They are about 1 order of magnitude lower than the values presented in Dobeš et al. (1986) – stressing the importance of regarding the occurrence or absence of fluids even more.

Table 1Petrological description of the stratigraphic layers of sediments in the Cheb Basin and the basement below, translated from Dobeš et al. (1986).

Download Print Version | Download XLSX

3.1 Large-scale ERT survey

The data acquisition was performed using the dipole–dipole configuration (AB MN, with A and B being the current injection electrodes and M and N being the potential electrodes), which is, considering the cost–effect relation for practical and theoretical reasons, most suitable for this large-scale ERT experiment. Transmitter and receiver units are physically separated on two lines reaching maximum dipole separations of 6.5 km (Fig. 1) while keeping the total length of required cables to a minimum as only neighboring electrodes have to be connected. Considering crop growth in June in this rural area and traffic by agricultural farming machines in general, other arrays are not effective with large cable spreads of several kilometers. Furthermore, we expected vertically oriented features (faults, “fluid channels”), as seen in previous studies (Nickschick et al., 2015), supporting the choice of using a dipole–dipole setup and achieving good results in previous studies at different locations with a similar setup (Flechsig et al., 2010; Pribnow et al., 2003; Schmidt-Hattenberger et al., 2013).

The experiment setup included 59 transmitter and voltage dipole locations by using 150 m dipole lengths in the outer (10 dipoles in the western and 11 in the eastern part of the profile) and 100 m length in the central part. While the receivers are stationary at fixed places during the campaign, the transmitter with the source dipole is moved to the feeding positions. Since the profile crosses streets and rural roads, small gaps needed to be left for current injections and voltage registrations, leading to a total number of 54 voltage reading positions and 47 current injections. To determine the horizontal position, we used a handheld GPS (Garmin GPS map 62s) with an accuracy of about 3 m. Elevations were then taken from a high-resolution digital elevation model. Two high-power transmitters (10 kW SCINTREX TSQ-4 and a self-developed 40 kW power transmitter) were used to inject a square-wave signal with an 8 s signal period and 50 % duty cycle (2 s positive, 2 s off, 2 s negative, 2 s off) using at least six cross-shaped, stainless-steel metal rods (1.5 m long) for grounding. For a total length of 20 min, current was injected. For 15 min of each 20 min injection period (112 total periods), we injected with the highest current possible, resulting in clear signals even at distances of several kilometers, and 5 min (37 total periods) with reduced power in case of overloads at nearby data loggers. The maximum injection current into the ground was 22.4 A with an average of 10.2 A for all injections. As voltage electrodes, non-polarizable electrodes (Ag–AgCl and Cu–CuSO₄) were used to avoid polarization effects over the current injection time. To register voltages, two data recorder types were used (24 RefTek Texan-125A single-channel recorder and 10 self-developed remote-controlled three-channel data loggers; Oppermann and Günther, 2018). A continuous registration of the full time series with a 100 Hz sampling rate for the single-channel recorder and 200 Hz sampling rate for the three-channel data logger was carried out during the survey to account for possible high-frequency noise signals. The field experiment is followed by comprehensive data preprocessing, including data storage, compilation of the raw data in a database system, raw data quality analysis, and raw data processing.

3.2 Small-scale ERT survey

In preparation for the large-scale experiment, several near-surface surveys using a commercial GeoTom multi-electrode device were carried out in proximity to the large profile. Due to the specific setup of the large-scale experiment and the limited resolution within the first tens of meters, additional surveys with small electrode spacings provide useful information about the near-surface resistivity. A total of 100 steel electrodes with a spacing of 5 m were used in these surveys, resulting in a total length of 495 m for a single profile. The setup is similar to the ERT profiles shown by Nickschick et al. (2015) and Nickschick et al. (2017) for comparison purposes. Thus, we also measured in Wenner alpha and Wenner beta configuration due to the good results from these previous studies. Both arrays have been combined and were inverted with the BERT software (see Sect. 3.4) using a vertical-to-horizontal smoothness factor (Coscia et al., 2011) of 0.2, i.e., making vertical gradients 5 times more sensitive than horizontal ones.

3.3 Processing of the large-scale ERT data

Natural and anthropogenic sources and industrial facilities near the profile lead to noise within the acquired voltage time series. To reduce noise and eliminate unwanted signals, data processing is required. This issue was addressed by a signal enhancement procedure with a selective stacking approach from Friedel (2000). The approach aims at stacking the acquired voltage time series U(t) (Fig. 3a) into separate cycles.

The first step in the processing procedure is a drift correction to remove the DC voltage parts and long-periodic drift components (Fig. 3b). This is realized by applying a filter function, yielding the drift-corrected function U_dr(t) that subtracts the moving mean value of the time series U(t) with a window size of the injection signal period M from the original time series U(t), as suggested by Friedel (2000):

This provides correct results in the case of a symmetric signal with an identical positive and negative amplitude, which is given in this case by controlling the source and assuming that the signal is not distorted by having a very high signal-to-noise ratio. The next step is to reduce short-term noise. In this case, this is done by stacking the events using the α-trimmed mean stack (Naess and Bruland, 1979; Friedel, 2000; Oppermann and Günther, 2018), in which every sample within the stacked signal period is sorted by amplitude, and the smallest and largest amplitudes that exceed a portion of α are rejected. Here, we used a rejection rate of α=10 %, resulting in a mean that is less susceptible to outliers by removing the most deviating 10 % of the samples. To determine the phase shifts between the injection signal and registered signal, a cross-correlation between the stacked signal and an ideal waveform needs to be found. This is done by stacking at an arbitrary point and determining the phase of maximum cross-correlation. As a final step, the response time of the current switching (transients) before reaching the plateau has be considered. A window is selected that ignores a fixed number of samples (typically 10 %) before and after the current switch. In the end, we get a stacked signal as seen in Fig. 3d. The voltage U is the half-difference between the positive (U_p) and negative (U_n) plateau voltages,

This has to be done for each of the 54 receiver dipoles at the 47 current injections, leading to a theoretical number of 2538 current–voltage pairs for this setup. However, this is reduced to a number of 2397 because voltage is not measured at the current electrodes.

Figure 3Processing steps of time series as an example. (a) Raw time series U(t), (b) time series U_dr(t) after drift correction, (c) stack distribution after cross-correlation, and (d) mean stacked signal with rejection windows to delete current switch effects (greyish areas) with positive (U_p) and negative (U_n) mean plateaus.

Download

In theory, every combination of current and voltage dipole is measured twice by taking into account the principle of reciprocity, which states that voltage and current can be interchanged. By comparing the apparent resistivity for forward values (AB dipole ahead of MN), ${\mathit{\rho }}_{\mathrm{f}}^a$, with the backward (AB behind MN) values, ${\mathit{\rho }}_{\mathrm{b}}^a$, one can compute the relative reciprocity error as

for each reciprocal pair. This value should be zero, but in practice it is not due to (i) different coupling of current injection fields compared to potential electrodes and (ii) individual noise levels at different voltage gains leading to different signal-to-noise levels. Therefore, it can be used as a measure of data consistency and also to derive error models (Udphuay et al., 2011) but only if a statistically large number of data are available.

Figure 4a shows the raw apparent resistivity ρ_a cross-plot as a function of current and voltage dipoles, which should be theoretically symmetric. White areas are blank due to injections at the respective voltage reading positions (three inner diagonals), dominant noise in the time series, or a missing cable connection. In the few cases in which the voltage was too high (e.g., at neighboring dipoles), the smaller current injection was chosen to fill in the missing data. In all other cases, the injection with higher currents leads to better signal-to-noise ratios.

Figure 4Raw data (all retrieved AB–MN pairs) as a function of current AB and potential MN dipoles. (a) Apparent resistivity (log ρ_a; Ω m) and (b) relative reciprocal error between forward (starting with current dipole 1∕2) and reverse measurements (%).

Download

Many factors interfere with the experiment and the voltage readings, decreasing the amount of reliable data. Strong, irregular signals of 16.7 Hz superimpose the data record of the westernmost logger (1∕2), which can be attributed to rail traffic 800 m south of the western part (Fig. 1) of the profile, leading to a high artificial signal input in general. The easternmost voltage readings (58∕59 and 59∕60) are often overlain by anthropogenic signals from the village of Kaceřov. Furthermore, the current injections show a highly disturbed injection signal, which we attribute to a buried gas pipeline, as indicated by their appropriate sign in the vicinity. Therefore these data had to be removed. Some of the planned injection dipoles (35∕36 to 38∕39 and 47∕48 to 48∕49) could not be accessed with the trailer-mounted current source due to roadside ditches and high crop growth at that time. Fortunately, the missing data (white columns) are mainly available through their reciprocals.

The reciprocal error is displayed in Fig. 4b. A large portion of the area appears grey; i.e., forward and backward data agree very well. For some data with short spacing (near the diagonal) the values deviate from zero due to different coupling. In general, reciprocal errors increase with increasing dipole separation and reflect the decreasing signal-to-noise ratio as a result of the strongly decaying signal strength.

In order to appraise the quality of the partially redundant sub-datasets, we used both data types in preliminary inversion runs with default parameters. It turns out that the root mean square (RMS) misfit of the upper right triangle was explicitly lower (11 %) compared to the lower left triangle (27 %), which shows more systematic structures in the misfit plot. Therefore, we decided to fill in missing data in the former with the latter. The further workflow has the aim of generating a homogenized pseudo-section. It consists of the following steps (see Oppermann and Günther, 2018):

• removing bad data (single outliers visible as points or point groups such as the aforementioned AB pair 44∕45);

• filling the missing values in the upper right triangle with the corresponding data in the lower left triangle;

• computing the data reciprocity for the doubled data from the resistivity; and

• replacing the corresponding resistances by the current-weighted mean of the two.

As a result, we obtain an apparent resistivity pseudo-section as known from multi-electrode measurements (Fig. 5), i.e., plotting the value as a function of the midpoint position and the separation (dipole distance normalized by dipole length).

Figure 5Unified dataset as apparent resistivity pseudo-section: measured data (a) and model forward response (b).

Download

For small separations, we observe low values (5–20 Ω m) in the west and higher values (40–200 Ω m) in the east. The apparent resistivity increases with separation, which is more pronounced in the western part. There are still two white stripes for a dipole with a missing registration.

3.4 Modeling and inversion of the resistivity data

The aim of the inverse modeling is to find a subsurface resistivity distribution that is able to reproduce the measured data. We use a smoothness-constrained Gauss–Newton inversion (Günther et al., 2006) implemented in the freely available software BERT (Günther and Rücker, 2019). The whole data processing and visualization uses the pyGIMLi framework (Rücker et al., 2017) in Python. The subsurface is discretized by triangles so that the measured topography can be taken into account accurately. The maximum model depth is determined by 1-D sensitivity analysis with about 130 m for the small profiles and 1300 m for the long one.

In the inversion process, the individual data points are weighted by error estimates consisting of a percentage error and an absolute voltage error so that measurements with lower voltage gain have less importance than those with strong signals. Reciprocal data can be analyzed statistically in order to obtain numbers for this error model (Udphuay et al., 2011). In our case, we determined a percentage error of 5 % and a voltage error of 2 µV, leading to maximum error estimates of 20 % for the large-scale ERT weakest signal at maximum distance. For the small-scale ERT profiles, no reciprocal data were available so we used the default value of 3 % plus 100 µV.

For the regularization, we used smoothness constraints of 1st order as described by Günther et al. (2006). However, to account for predominantly layered structures (larger correlation length in the x direction compared to the z direction), we applied a vertical smoothness factor (see Coscia et al., 2011) of 0.1; i.e., purely vertical gradients in the model are 10 times less penalized than purely horizontal gradients. The overall regularization parameter (300) was chosen such that the data were fitted within the estimated noise level, i.e., with a chi-square error (root mean square of error-weighted misfit) of about 1. Whereas this corresponds to RMS values of about 5 % for the short profiles, the large profile shows a relative misfit of about 12 %.

The forward response, i.e., the apparent resistivity theoretically measured over the retrieved resistivity subsurface, is displayed in Fig. 5. One can see that the main structures are reproduced by the model but not the detailed outliers due to error weighting, resulting in the overall misfit of 12 %.

3.5 Gravity survey

In conjunction with the resistivity survey, we also measured gravity along the ERT profile in order to have additional geophysical data for interpretation. For this purpose, a LaCoste & Romberg D-188 gravimeter was used for gravity surveys in 2017 along the ERT profile. Its resolution is 0.001 mGal, and we achieved an accuracy of 0.006 mGal. In the central part of the profile, very detailed measurements from the previous investigation of the Hartoušov degassing zone from 2012 on profile 2 from Nickschick et al. (2015) were included. To double-check the accuracy of the new surveys in comparison to the older one, several points from that profile were located and remeasured. The average difference was only 0.008 mGal. The spacing on the profile between each measurement station was 40–60 m, while the spacing in the central zone on this profile is denser (10–40 m). Thus, a total of 170 stations exists along the profile. The gravity measurements were referenced to the Czech national gravity network. All essential corrections were applied (drift, tidal, latitude, free-air, Bouguer, terrain). Coordinates were observed by Trimble R9 real-time kinematic (RTK) technology, and the accuracy of all these measurements was better than 0.03 m in the vertical component. Terrain corrections were calculated from an accurate digital elevation model (DEM) of 1 m resolution to the distance of 250 m, the outer part of the correction to 167 km from the SRTM90 DEM. As the profile was located in the Cheb Basin, the reduction density of 2300 kg m⁻³ was applied to compute the Bouguer anomalies.

4.1 Small-scale ERT

Figure 6Resistivity distribution of the small-scale ERT profiles: 1 (a), 2 (b), and 3 (c) (z: m a.s.l.; see Figs. 1 and 2 for locations and lithology). Transparency represents the coverage (cumulative sensitivity) and helps avoid the interpretation of uncertain model parts. The relative RMS values are 4.7 %, 17.4 %, and 4.5 %, respectively.

Download

The three short ERT profiles (625–700 m long) provide insight into the uppermost (approximately 100 m) resistivity distribution along the large-scale profile (Fig. 6). We chose an identical color scale for all results that helps comparison with the large-scale profile. We used the coverage (cumulative sensitivity of the final model) for the alpha shading of the inversion results so that poorly covered model regions will not be interpreted. We refer to Ronczka et al. (2017), who substantiate the use of coverage as a simple approximation for the model resolution by using synthetic and field ERT data.

Profile 1 (Fig. 6a; final RMS after three iterations is 4.7 %), located in the western part of the large-scale profile, reveals that the first 5 m of this profile feature resistivities of less than 100 Ω m. This layer is on top of a rather massive and homogeneous compound of conductive rocks that is characterized by resistivities of 15–60 Ω m between 5 and 20 m of depth and an even more conductive (<15 Ω m) zone beneath. This resistivity distribution encountered here fits into the geological description of drilling B-18. The first few meters consist of resistive Quaternary sand and loam compared to the lower resistivity that is the underlying Cypris formation. The drill log describes the area beneath 20 m as water-saturated, so it can be assumed that the first 20 m are not saturated and thus slightly less conductive.

Profile 2 (Fig. 6b; final RMS after nine iterations is 17.4 %), crossing the mofette field Hartoušov, confirms the findings from Nickschick et al. (2015): a resistive (> 150 Ω m) layer of ca. 15 m thickness can be measured on top of the more conductive zone. At approximately 400 m of profile distance, just as the elevation increases towards the east, a significant thickening of the high-resistivity near-surface layer can be observed. The resistivity distribution in the western part of profile 2 fits the description of drilling SA-30 and the new drilling HJB-1 (Bussert et al., 2017): the first 15 m consist of gravel, sand, and peat, resulting in overall higher resistivities compared to the Tertiary sediments below. Discrepancies in the core description between drills SA-30 and HJB-1 reveal that deposits (clay and gravel) from the Vildštejn formation are found in the area. We link the sudden shift in resistivity and elevation from 400 m onward to the increased thickness of the Vildštejn deposits towards the east, as stated by the drill logs. This sudden and sharp lithology shift is linked to the course of the PPZ and vertical offsets of a few tens of meters due to various stages of subsidence and lifting (Bankwitz et al., 2003a; Peterek et al., 2011; Kämpf et al., 2013; Rojik et al., 2014; Nickschick et al., 2015). It is to be noted that the vertical plume-like anomalies could be linked to areas of strong CO₂ degassing at the surface as reported in previous studies (Flechsig et al., 2008; Nickschick et al., 2015, 2017).

Profile 3 (Fig. 6c; final RMS after 14 iterations is 4.5 %) reveals a 10–15 m thick layer with resistivities above 300 Ω m on top of a massive compound of rocks with about 150 Ω m, which is significantly higher than in profiles 1 and 2. At about 100 m of depth, resistivity decreases, but this represents the investigation depth limit. Core descriptions from nearby drills, such as B-1 and SA-31, indicate a 10–12 m thick layer of Quaternary deposits as the topmost layer. Clayey and silty–sandy Vildštejn deposits, however, have reached thicknesses of 60–80 m in this area according to the core descriptions, which reflects higher resistivities compared to the very conductive Cypris formation at the bottom.

4.2 Large-scale ERT profile

Figure 7 shows the inversion result of the long profile after four iterations (RMS evolution – 81 %, 29.9 %, 18.5 %, 14.6 %, and 14.5 % as the final RMS). On top, the lithology provided by the neighboring drills is plotted as colored box columns, indicating the limited depth that has been achieved by the drills.

Figure 7Inversion result (resistivity distribution) of the large-scale profile with the lithology columns of the boreholes (a) and the Bouguer gravity (b) (z: m a.s.l). Colors for each stratigraphic unit are identical to Fig. 2: green – Vildštejn formation, light blue – Cypris formation, brown – coal, red – Lower Sand formation, pink – phyllitic–granitic basement. The relative RMS is 14.5 %, and coverage-based alpha shading is used as in Fig. 6.

Download

The 2-D resistivity distribution of the profile shows remarkable differences in the structural composition in the western half of the profile compared to the eastern half. We observe a highly conductive layer of <30 Ω m of about 200 m thickness above a basement of higher resistivity (>100 Ω m) in general. The transition is gradual. At about 2500–3000 m along the profile, these layers dip towards the east and form a trough-like structure before ascending again upwards to the eastern end. This also leads to the occurrence of another layer of >100 Ω m at the surface between 3200 and 5800 m, which reaches a maximum thickness of about 300 m. The lowest resistivities (5 Ω m) are found along 4000–5000 m along the profile at a depth of 300–500 m.

4.3 Gravity

The gravity survey (Fig. 7c) reveals a total maximum relative gravity difference of about 9 mGal along the profile between the local maximum at ≈1500 m and the minimum at 6300 m. It is to be noted that the gravity minimum is measured at the point of highest elevation. The maximum is located where a high-resistivity anomaly is observed in the profile, and the minimum is slightly west of the area where the lowest resistivities were measured. The slight shift between these two observations might be related to the different sensitivity of the electric resistivity and density towards changes in the lithology in north or south of the profile. This gravity trend is enhanced by the W–E-trending contact of phyllitic (on the southern side) and granitic (on the northern side) rocks in the basement, according to Hecht et al. (1997). The central section around the Hartoušov mofette field is located on the crossing of this zone with the Počátky–Plesná fault zone and the gravity gradient delineating the deepest part of the basin. Such tectonic and structural zones form permeable channels for deep fluids and have been mentioned before for this area (Bankwitz et al., 2003b; Kämpf et al., 2013; Bräuer et al., 2008; Fischer et al., 2014, 2017). In Nickschick et al. (2015), we proved that detailed microgravity measurements in the mofette area are capable of locating particular small-scale degassing channels due to decreased bulk density of the rocks, which are in the range of a few tens of microgals and thus not visible on this scale. At the eastern end of the profile, a gravity increase indicates the contact of sediments with outcropping basement of the Krušné hory Mountains.