Interdisciplinary Fracture Network Characterization in the Crystalline Basement: A case study from the Southern Odenwald, SW Germany

. The crystalline basement is considered a ubiquitous and almost inexhaustible source of geothermal energy in the Upper Rhine Graben and other regions worldwide. The hydraulic properties of the basement, which are one of the key factors for the productivity of geothermal power plants, are primarily controlled by hydraulically active faults and fractures. While the most accurate in situ information about the general fracture network is obtained from image logs of deep boreholes, such data are generally sparse, costly and thus often not openly accessible. To circumvent this problem, an outcrop analogue study 5 with interdisciplinary geoscientific methods was conducted in the Tromm Granite, located in the southern Odenwald at the northeastern margin of the URG. Using Light detection and ranging (LiDAR) scanning, the key characteristics of the fracture network were extracted in a total of five outcrops, additionally complemented by lineament analysis of two different digital elevation models (DEMs). Based on this, discrete fracture network (DFN) models were developed to calculate equivalent permeability tensors under assumed reservoir conditions. The influence of different parameters, such as fracture orientation, 10 density, aperture and mineralization was investigated. In addition, extensive gravity and radon measurements were carried out in the study area, allowing to map fault zones with naturally increased porosity and permeability. Gravity anomalies served as input data for a stochastic density inversion, through which areas of potentially increased open porosity identified. A laterally heterogeneous fracture network characterizes the Tromm Granite, with the highest natural permeabilities expected at the pluton margin, due to the influence of large shear and fault zones.


Introduction
The Upper Rhine Graben (URG) represents a region with a high potential for deep geothermal projects in Central Europe due to a significantly increased geothermal gradient of locally more than 100°C km-1 (e.g. Agemar et al., 2014). Often based on  (Klemm, 1900(Klemm, , 1928(Klemm, , 1929(Klemm, , 1933. Mapped  The digitized lineaments were used to calculate the lineament density P20 (number of fractures per unit area) and intensity P21 (total length of fractures per unit area) (Sanderson and Nixon, 2018).

Outcrop analysis 125
Five abandoned quarries located across the Tromm Granite were selected for detailed structural analysis of the fracture network ( Fig. 2). As also described in Bossennec et al. (2021), the RIEGL VG 400 LiDAR instrument was used to generate highresolution point clouds (point spacing ≤ 1 cm) of the outcrop walls. Compared to classical scanlines, this approach allows for relatively quick acquisition of large structural datasets. At the same time, the statistical bias is reduced as all visible fractures are detected, not only those that cross a 1D line. For an in-depth discussion of the reliability of LiDAR for outcrop analysis,  The raw LiDAR data were first imported into RiSCAN PRO to merge individual scans. Further analysis of the point clouds was performed using the open-source softwares CloudCompare and QGis. The point cloud was resampled to less than 2 million points to reduce the computational effort of the following steps. Afterwards, the orientation of the surface normals was calculated by triangulation between the points and converted to dip and dip direction. Based on this, the Ransac shape 135 detection plugin was applied to automatically extract the orientation of continuous fracture planes (e.g. Drews et al., 2018).
The following parameters were chosen for this step: maximum distance to plane = 5 cm, scanning distance = 20 cm, maximum normal deviation = 10°. Each detected plane was visually inspected and removed if it did not represent natural fractures.
Besides the automatic plane recognition, the LiDAR data were also manually interpreted in QGis to investigate the fracture length, density and connectivity (Fig. 3). For this purpose, side projections of the point clouds were rasterized and hill shade 140 maps were again generated. Visible fractures were then digitized to compute the fracture areal density P20 and intensity P21.
Additionally, the linear fracture frequency P10 was extracted along virtual horizontal scanlines for each outcrop. The topology of the fractures was furthermore studied to characterize the connectivity of the network. For this, the tips of all fracture branches were classified into three groups: isolated (I), abutting (Y) and crossing (X) nodes. The average number of connections per line c L was calculated from the number of nodes per type (Sanderson and Nixon, 2018). 145 The results of the lineament and outcrop analyses are finally summarized in a normalized trace length cumulative frequency plot with a power-law fitted to the data, which describes the relationship between frequency and the cumulative distribution of fractures lengths (Pickering et al., 1995;Marrett et al., 1999).

DFN modelling
DFN models were generated with the software FracMan to quantitatively model the hydraulic properties of the fractured crys-150 talline basement based on the structural parameters acquired in the field. Fracture orientations were implemented by performing a cluster analysis on the dip directions and dip angles extracted from the LiDAR data. The fracture density was defined along a virtual horizontal borehole using the calculated P10 values. The fracture length distribution was set according to the computed power law. A lower cut-off of 70 cm was applied, as significant censoring, i.e. under-representation of short fractures, occurs below this length. The effective fracture aperture largely governs the hydraulic conductivity of fractures. Due to exhumation 155 and weathering processes, measured aperture values at near-surface outcrops are usually not reliable (Place et al., 2016). Instead, an exponential distribution of the apertures and following Sausse and Genter (2005) three possible mean values (10 µm, 50 µm and 100 µm) were assumed. A more accurate approach would be to relate the aperture to the normal stress on the fracture plane (Bisdom et al., 2017), but as the local stress magnitudes are largely unconstrained in the Tromm Granite, this was not pursued further. Additionally, previous studies showed that a major part of the naturally occurring fractures in the 160 crystalline basement are mineralized at reservoir depth and therefore only a small share of the fractures allow fluid flow (Genter and Traineau, 1996;McCaffrey et al., 1999;Evans et al., 2005). For this reason, three different scenarios for the proportion of hydraulically active fractures in the DFN model were defined (1, 10 and 100 %).
For a sufficient number of discontinuities, the fractured basement behaves like an anisotropic porous medium. The equivalent porous medium (EPM) permeability tensor can thus be calculated for a DFN model by, e.g., the approach of Oda (1985). The 165 undisturbed rock matrix is considered impermeable (Jing and Stephansson, 2007;Weinert et al., 2020), implying that fluid flow occurs exclusively through connected fractures. The directional permeability is related to the size, orientation, opening and connectivity of the fractures. One key factor is the relationship between fluid flow along a fracture and the aperture, described by a cubic law (Snow, 1965). This relationship is based on the assumption of laminar flow between two parallel surfaces, which is often not the case due to the irregular surface and aperture of the fractures and can therefore lead to errors.

170
The permeability tensors were computed along a regularly spaced grid with a cell size of 10 m to reduce the computational effort and afterwards, mean values were calculated for the entire DFN model. Different cell sizes between 1 m and 20 m were tested, revealing no significant differences in the resulting mean permeability tensor.

Geophysical surveys
3.2.1 Gravity data acquisition 175 During two surveys in summer 2020 and spring 2021, gravity measurements at 431 stations along 11 profiles have been conducted in the Tromm Granite (Fig. 2). Since a differential GPS was used to determine the position, the campaigns were restricted to the southern, less densely forested part. The GPS data were corrected against known fix points, resulting in c. 10 to 20 cm vertical accuracy. Gravity measurements were performed using the Scintrex® CG-6 Autograv gravimeter with an average station spacing of 100 m respectively 20 to 25 m close to presumed fault zones. Base measurements were taken three 180 times per day at a fixed station to record the instrument drift. Measurements with a standard deviation greater than 0.05 mGal were excluded. A complete Bouguer anomaly was calculated for all gravity stations by applying the standard correction density of 2,670 kg m −3 , which corresponds approximately to the mean rock density of the Tromm Granite (Weinert et al., 2020), as also confirmed by Nettleton analysis (Nettleton, 1939). Particular focus was on the topographic correction, which along some profiles reaches up to 2 mGal due to the steep terrain. The calculation was performed with the software GSolve (McCubbine km, DEM 1 km). Taking into account all uncertainties in the data acquisition and processing, especially the height error and the standard deviation of the measurement, a cumulative error of the Bouguer anomaly of less than 0.1 mGal was determined. For the regional gravity signal analysis, c. 5300 additional data points provided by the Leibniz Institute for Applied Geophysics (LIAG) and the Hessian Administration for Land Management and Geoinformation (HVGB) were used within a radius 190 of 50 km around the survey area. Together with the newly acquired data, a Bouguer anomaly map with a nominal resolution of 20 m was calculated using the minimum curvature interpolation method. A series of high-pass filters with cut-off wavelengths of 10 km, 5 km, and 2 km was then applied to subtract the regional gravity field.

Inversion of the gravity data
A stochastic 3D inversion of the high-pass filter Bouguer anomaly (10 km cut-off wavelength) was performed to infer the 195 density distribution and the porosity in the subsurface. The commercial platform GeoModeller (Intrepid Geophysics) was used for this purpose, which employes a Monte-Carlo Markov-Chain algorithm to invert geophysical data. A detailed discussion of the methodology is available in previous studies (Guillen et al., 2008). The model domain has an extension of 7 km in E-W and 6 km in N-S direction and a depth of 2 km. The upper boundary is defined by the 10 m DEM. Given the relative homogeneity of the pluton with respect to the matrix density and the lack of structural input data, an unconstrained inversion was performed.

200
The continuous model was converted into a discrete cuboid voxel model with a cell size of 50 x 50 x 50 m. Further decreasing the cell size potentially resolves more details but exponentially increases the computational time of the inversion. As starting value for the rock density, a mean density of the Tromm Granite of 2,670 ± 50 kg m −3 was defined (Weinert et al., 2020).
The algorithm first calculates the geophysical effect of the starting model, in this case a homogenous half-space, and then uses a Bayesian approach to determine the likelihood of the model. In subsequent iterations, random variations of the model 205 are generated according to the probability distribution of the rock density. Models that lead to a reduction in the misfit between calculated and measured gravity anomalies have a higher likelihood and are stored. After 250 million iterations, a larger collection of possible models is generated, allowing statistic evaluation. Finally, the porosity is estimated assuming the above mentioned homogeneity of the Tromm granite bÿ Where ρ bulk is the bulk density, ρ f luid the fluid density (c. 1,000 kg m −3 ), ρmatrix the matrix density and Φ the porosity.
Note that this equation can not simply be applied in the southernmost Tromm Granite, where significant lithological heterogeneity is observed, resulting in variations of the bulk density without the influence of increased fracture porosity.

Radon measurements
Radon is a naturally occurring radioactive gas that is concentrated in the soil air. The most abundant Rn-isotope with a propor- 2000Pro at 20 points on one profile that crosses two presumed fault zones (Fig. 2b). Soil air was sampled using a hollow probe driven 1 m deep into the subsurface. A pump was connected to this probe, which flooded the ionization chamber of the radon monitor. The activity concentration was measured within 1-minute cycles. After 15 minutes, the air in the chamber was completely exchanged. The pump was then switched off, and the chamber short-circuited for an additional 10 minutes.
During this time interval, most of the very short-lived Thoron (Rn-220, t1/2 = 55.6 s) decayed so that the final reading corre-225 sponded only to Rn-222 concentration. Furthermore, soil samples were taken with a slotted probe at all stations to determine the soil type. Repeated measurements were performed at a base station to quantify the temporal variability of the concentration measurements.  Table 1 compare the faults that were determined by geological mapping in the Tromm Granite area with the geologically significant lineaments extracted from the SRTM model and the DEM with 1 m resolution. A total of 30 faults with a characteristic length (arithmetic mean of the fracture length) of c. 2,900 m were extracted from geological maps (modified after Klemm, 1900Klemm, , 1928Klemm, , 1929Klemm, , 1933HLUG, 2007), corresponding to a P21 value of 0.0014 m m −2 . The total number of 235 elements that were identified with the lineament analyses is significantly higher (177 for SRTM and 471 for the 1m DEM).

Fracture network properties
Their characteristic length of 1187 m and 680 m, respectively, is smaller and decreases with increasing resolution. The P21 is 0.0034 m m −2 for the SRTM lineaments and 0.0051 m m −2 for the lineaments of the high-resolution DEM. In all data sets, a heterogeneous spatial distribution of the faults or lineaments can be observed. The element density is highest in the eastern part of the Tromm Granite, i.e. in the area influenced by the Otzberg Shear Zone. The density is significantly lower in the west, 240 especially for the mapped faults and the SRTM lineaments.
The main strike of the mapped faults ranges from N160°E to N170°E, which corresponds approximately to the direction of the maximum horizontal stress s Hmax (Reiter et al., 2016). In contrast, the main set of SRTM lineaments strikes with N090°E ±30°. The strike directions of lineaments from the high-resolution DEM show nearly an equal distribution, with a slight accumulation of lineaments at N100°E. Granite. Remnants or intrusions of other rock species are scarce, but veins of younger granites can frequently be observed.

250
These two outcrops have the lowest areal fracture intensity with a P21 of 2.43 m m −2 and 2.83 m m −2 , respectively (Table   1). Conversely, the fracture characteristic length is the longest here, reaching 1.28 m and 1.62 m respectively. The main set of fractures dips steeply and strikes N160°E ±20°, which corresponds to the main direction of the geologically mapped faults in the Tromm Granite (Fig. 4). A second, subordinate set of steeply dipping conjugate fractures strikes N060°E ±10°. Shallow dipping fractures are very rare.

255
The most extended outcrop examined is an abandoned quarry with dimensions of c. 150 × 250 m close to the village of Ober-Mengelbach at the southern border of the Tromm Granite (Fig. 3). In contrast to Borstein and Streitsdölle, the lithological conditions found here are more heterogeneous. Meter-to ten-meter-wide amphibolite zones were observed throughout the quarry that are highly deformed and intruded by granite or granodiorite. The magmatic contacts are usually not abrupt but rather characterized by mixed forms of amphibolite and granitoids. Again, several generations of granite intrusions can be 260 Figure 4. Summary of the lineament analysis in the Tromm Granite area: (a) compilation of mapped faults from various geological maps (modified after Klemm, 1900Klemm, , 1928Klemm, , 1929Klemm, , 1933HLUG, 2007); (b) regional analysis using SRTM data with 1 arcsecond resolution (van Zyl, 2001); (c) local analysis using 1 m DEM provided by the HVBG.
distinguished. The distribution of fractures was investigated along four 2D profiles with a length between 20 and 30 m, and a height of 10 m (see Fig. 2 for location). The P21 ranges from 3.60 to 5.87 m m −2 and is thus about twice as high as in the central Tromm Granite. The extraction of fracture orientations using the Ransac filter was carried out for all outcrop walls to obtain the most comprehensive data set possible. Again, the primary set of fractures strikes N160°E ±20°and a secondary set strikes N070°E ±10° (Fig. 5).

265
The two smaller outcrops in Hammelbach and the Weschnitz valley are located at the northeastern border of the Tromm Granite. Here, a fine-to medium-grained, cataclastic granite is predominant, which was considerably affected by the adjacent Otzberg Shear Zone. Consequently, the P21 is the highest with 10.82 m m −2 and 9.07 m m −2 , respectively. The fracture orientation also differs significantly from the other three locations. In Hammelbach, the fractures strike almost exclusively N100°E ±20°. In the Weschnitz Valley at the northern margin of the Tromm Granite, two fracture sets were found, striking 270 N050°E ±10°and N130°E ±20°, respectively. These directions correlate well with the orientation of the close-by lineaments.   Granite in terms of fracture density (Fig. 7). The estimated EPM permeabilities in x-(E-W-), y-(N-S-) and z-direction are  (1900,1928,1929,1933) and HLUG (2007). summarized in Table 2. The fracture mean aperture has the largest influence on the permeability, as these two parameters are related via a cubic law. Practically speaking, this means that increasing the aperture by one order of magnitude leads to three orders of magnitude higher permeability. The proportion of open fracture, in contrast, is linearly related to permeability, i.e. a 295 tenfold increase also increases permeability by a factor of ten. The orientation of the fracture sets has furthermore a significant effect on the permeability of the basement. At Borstein, the permeability in the main direction of the fractures (k yy ) is almost one order of magnitude higher than perpendicular to it (k xx ). Finally, the difference between Borstein and Weschnitz Valley reaches again up to one order of magnitude, depending on the direction, for the same aperture and proportion of open fractures.

Length distribution and fracture connectivity
This variability is mainly due to approximately four times higher fracture density in the second outcrop (Fig. 8).

300
To test the transferability of the results to crystalline reservoirs in the URG, a comparison with hydrogeological data e.g. from Soultz-sous-Forêts is useful. Here, the mean permeabilities of the fractured granitic basement range from 1E-16 to 1E-14 m 2 at reservoir depth of 3 to 5 km (Vogt et al., 2012;Baujard et al., 2017;Egert et al., 2020). Accordingly, realistic permeabilities result for (1) a mean aperture of 10 µm and (2)  It should be noted that the hydraulic properties of fractured reservoirs are subject to strong spatial variations. For example, permeability can be increased by several orders of magnitudes close to active faults. In contrast, at larger distance from these faults or large-scale fractures, the mean permeability of the basement is rather in the order of 1E-18 to 1E-17 m 2 .   The gravity field in this area is dominated by an NW-SE oriented regional trend (Fig. 9b) that was obtained by applying a low-pass filter with cut-off wavelength of 10 km. Residual anomalies 315 involving presumably lower depth ranges were obtained by applying high-pass filters with decreasing cut-off wavelengths from 10 to 2 km (Fig. 9c-e). The residual field exhibits distinct positive and negative anomalies. However, especially in the central part of the Tromm Granite, the lack of data points leads to considerable uncertainties.

Gravity and Radon Anomalies
The strongest positive anomaly of 1 to 1.5 mGal is located north of Wald-Michelbach and coincides with a major lineament.
Similarly, a positive anomaly of 0.5 to 1 mGal can be observed along the presumed fault zone between Zotzenbach and Wald-320 Michelbach. The strongest negative anomaly with an amplitude of c. -0.5 to -1 mGal extends over several kilometres from SSW to NNE at the western boundary of the Tromm Granite to the Weschnitz Granodiorite. Note that this area is covered by locally more than 20 m but generally less than 10 m thick Quaternary sediments. However, the anomaly increases and persists with increasing cut-off wavelength. Another negative anomaly of c. -0.4 mGal is located at the eastern boundary of the pluton, southeast of the village of Tromm. Here, the granite is highly fractured due to the proximity to the Otzberg Shear Zone, which 325 is also indicated by the high concentration of local lineaments. However, a direct correlation between Bouguer anomalies and individual faults or lineaments is usually not observed.
Besides these larger anomalies, short-wavelength variations of the gravity signal in the range of -0.3 to 0.3 mGal occur on individual profiles, which is still significantly higher than the cumulative uncertainty of the Bouguer anomaly.

330
Results of the stochastic gravity inversion are shown as differences to the homogenous density of 2,670 ± 50 kg m −3 of the starting model in Fig. 10.

Comparison of gravity and radon measurements
A comparison of the radon activity concentration in soil air with the corresponding Bouguer anomalies is shown in Fig. 11 (see Fig. 2 for the location of the stations). A background activity of c. 25 kBeq m −3 was determined. The repeated base measurements furthermore revealed a standard deviation of 5 kBeq m −3 . Near the two assumed fault zones, a significant increase in the activity concentration can be observed with two pronounced peaks of 5 to 7 times the background value. The high-pass filtered Bouguer anomaly with 2 km cut-off.
peak is located close to the northeastern fault zone, but here, as with the southwestern fault, a positive Bouguer anomaly is present. Thus, there is only a partial correlation between the two data sets.

360
Like the fracture length, the connectivity of the fracture network seems to be independent of scale or location. All outcrops and lineament maps indicate a dominance of Y-nodes, which is in clear contrast to the northern Odenwald, where I-and Xnodes represent the largest share (Bossennec et al., 2021). This can be attributed to different regional tectonic conditions during the intrusion, cooling and exhumation, or to overprinting under variable stress conditions.
Compared to fracture length and connectivity, the orientation of the fracture sets shows some scale-dependent and spatial 365 variations. In the outcrops Ober-Mengelbach, Borstein and Streitsdölle, the fracture orientations are controlled by the main fault direction of N160°E ±20°E in the Tromm Granite. In contrast, the fracture sets of the two outcrops Hammelbach and Weschnitz Valley are more influenced by local fault zones, as indicated by the lineament analysis (Fig. 4). Furthermore, the N-S trend of the mapped faults can hardly be found in the two lineament maps. Instead, elements perpendicular to it, i.e. oriented E-W, are dominant here.

370
Similar to the fracture orientation, the fracture density is subject to considerable lateral changes, which can be attributed to the influence of large-scale tectonic structures, especially at the pluton margins. In the eastern part of the Tromm Granite, the basement is deformed by the nearby Otzerg Shear Zone. As a result, there is an evident accumulation of lineaments and the outcrops show by far the highest fracture density. Medium fracture densities were found in Ober-Mengelbach, in the southern part of the pluton, i.e. at the border to the Schollenagglomerat. Although this area lacks pronounced long fault zones, the 375 lithological heterogeneities led to a more intense granite deformation than in the central Tromm Granite. Accordingly, the lowest fracture density was found in the outcrops Borstein and Streitsdölle.
In summary, the Tromm Granite is likely not characterized by a complete fractal fracture network, which is consistent, e.g., with studies from the western rift shoulder (Bertrand et al., 2018). Although fracture length and connectivity seem to be independent of scale and location, orientation and fracture density show variations with location. This fact must be considered 380 when evaluating and modelling the basement, as varying fracture orientation and density can increase or decrease permeability by up to one order of magnitude depending on the assumed mean apertures (Fig. 8).

Interpretation of gravity and radon anomalies
The measured gravity anomalies provide insights into the subsurface density distribution of the Tromm Granite (Figs. 9 and 10).
Negative anomalies of up to 1 mGal are concentrated at the western and eastern boundary of the pluton, where the basement 385 is strongly deformed and fractured. Comparable results in the Argentera Massif (NW Italy) were linked to a few percent of fracture porosity (Guglielmetti et al., 2013). However, several tens of meters thick layer of low-density Quaternary sediments or a thick basement weathering horizon could also lead to negative gravity anomalies. Borehole data from the HLNUG suggest that Quaternary sediment thicknesses typically do not exceed 10 to 15 m, accounting for a maximum of -0.2 mGal of the signal.
Nevertheless, it is known that the weathering zone in the granite is locally 20 to 40 m thick, which could account for a gravity 390 anomaly of up to -0.5 mGal.
In general, individual faults can rarely be accurately traced with the acquired gravity data in the Tromm Granite, as the influence of fracture porosity on bulk density is too small. Instead, areas can be identified where a high density of faults and fractures lead to increased porosity and thus to significant density reduction. Accordingly, the Tromm Granite is potentially structurally weakened at the contact with the Weschnitz Pluton in the western part of the study area. Unfortunately, there are 395 neither larger outcrops nor well data available, leaving this assumption speculative. The slightly smaller negative anomaly at the eastern boundary to the Buntsandstein can be explained by the proximity to the Otzberg Shear Zone. Here, the pluton is presumably characterized by similar structural properties as in the Hammelbach, and Weschnitztal outcrops, which means that the fracture density and thus the porosity are increased. Interestingly, the anomaly does not extend over the entire damage zone at the eastern margin of the Tromm Granite but is concentrated in a limited area with a high density of intersecting lineaments.

400
A possible explanation is that the fractures are partially mineralized with e.g. barite (e.g. Tranter et al., 2021), resulting in increased bulk density.
Positive gravity anomalies of up to 1.5 mGal can be observed at the southern Tromm Granite along two fault zones. In Gadern, several lamphropyric intrusions were mapped and, as in the quarry of Ober-Mengelbach, localized amphibolitic zones are present. These mafic rocks have a considerably higher density (2700 -3100 kg m −3 ) than the Tromm Granite, which 405 explains the gravity high. Due to the lithological heterogeneity in the southern Tromm Granite, quantification of fracture porosity using the gravity data is difficult here.
Radon measurements were carried out along just one profile due to the high time consumption of this method. Accordingly, a regional interpretation of the results is only possible to a limited extent. Nevertheless, the determined radon anomalies give helpful indications about the architecture of the analyzed fault zones. Two distinct radon peaks indicate localized permeable 410 fracture zones in the granite. The highest activity correlates with a negative Bouguer anomaly which further supports this assumption. Interestingly, the peaks are not located directly above the assumed position of the faults, but in the damage zone a few metres to tens of metres to the sides, suggesting low permeability in the fault core (Caine et al., 1996). The comparison of gravity, radon and structural data illustrates the complex architecture of fault zones in the crystalline basement (Faulkner et al., 2010;Bossennec et al., 2021Bossennec et al., , 2022, consisting of several adjacent permeable and impermeable zones, which may also 415 be influenced by clay mineralization. By combining gravity and magnetic field data, it is thus possible to study on the one hand the dimensions of the fracture network, and on the other hand to evaluate its behavior (sealed or open).

Implication for deep geothermal exploration and GeoLaB
The Tromm Granite represents a suitable site for the planned geothermal underground research laboratory (GeoLaB), as the main criteria proposed by Schill et al. (2016) are met. Firstly, except for the southern part, the pluton exhibits low geological 420 complexity with a rather homogeneous crystalline matrix. The high fracture density of 1.62 to 6.95 m −1 and good connectivity ensures sufficient hydraulic fracture permeability for the experiments. As required, the Tromm Granite is located in a normal faulting to strike-slip regime. In addition, the primarily NNW-SSE oriented fractures have a high reactivation potential in the ambient stress system. There is no extensive drainage in the area due to historical mining activities in adits, resulting in controllable hydraulic boundary conditions. Finally, the topography in the central part of the Tromm Granite can be described 425 as plateau-like, allowing an overburden of the tunnel between 300 and 400 m, which assures undisturbed stress conditions. The interdisciplinary data set described herein will serve as an important basis for the planning of further exploration activities in the area as well as the final site selection.
Furthermore, the Tromm Granite is a well-suited outcrop analogue for the crystalline basement in the URG. The granitic body has a similar mineralogical composition as the reservoir rocks e.g. in Soultz-sous-Forêts or Rittershoffen (Traineau 430 et al., 1991;Dezayes et al., 2005a;. Granitoids are dominant in the northern URG, as inferred from the joint inversion of gravity and magnetics data (Baillieux et al., 2013;Frey et al., 2021b). Moreover, the Tromm Granite was intruded in the same tectonic setting and overprinted under comparable conditions as the granites of the URG. Nevertheless, a direct transfer to deep geothermal reservoirs, where in-situ hydrothermal alteration and mineralization as well as complex geomechanical processes occur, is challenging considering that weathering and exhumation significantly affect the near-surface 435 fracture network. Unfortunately, no major fault zone is exposed in the study area, making it virtually impossible to evaluate the true nature of mineralization and fluid circulation patterns. The applied gravity and radon surveys help to overcome this lack of outcrops and provide insights into the basement's permeability structure.
A DFN parameter study was carried out to estimate the hydraulic properties of the Tromm Granite under assumed reservoir conditions (Fig. 8). The calculated permeabilities show a range of several orders of magnitude, indicating the uncertainties of but the effect of individual parameters can be well followed. The fracture aperture primarily determines the permeability of the basement, which is in agreement with detailed sensitivity studies, e.g. performed by Niven and Deutsch (2009) or Mahmoodpour et al. (in review). The degree of mineralization (here considered as the proportion of open fractures), the fracture density, and the fracture orientation also influence the Oda permeability, but have a smaller effect. DFN modelling suggests 445 that only a small proportion of the total fracture network is contributing to flow. With an average aperture of 10 to 50 µm, probably only 1 to 10 % of the fractures allow fluid flow, which fits well with observations from Soultz-sous-Forets (Sausse et al., 2010;Egert et al., 2020). It should also be noted that k xx and k yy show a difference of up to one order of magnitude (e.g. Mahmoodpour et al., 2021), which is particularly relevant for planning the well path trajectories of geothermal doublets and the experiments in GeoLaB. Accordingly, a geothermal doublet oriented approximately in a direction, that the open-hole 450 sections intersect a high number of N-S oriented fractures probably yields the highest production rates, which is consistent with observations from e.g. Rittershoffen.
Minimizing induced seismicity during stimulation and operation represents a major challenge for deep geothermal exploitation of the crystalline basement (Zhang et al., 2013;Meller and Ledésert, 2017;Rathnaweera et al., 2020), and will be addressed experimentally in GeoLaB. Stimulation is generally more feasible in reservoirs with naturally elevated permeability, because 455 lower injection pressures are required. The highest permeability is expected near large-scale fault zones with well-developed damage zones and hydrothermal overprinting. However, the structural geological investigations in Tromm Granite show that in certain areas sufficient permeability may also occur in the outer damage zones of large faults, i.e. at a distance of several hundred meters to kilometers from the fault core.
Apart from hydrogeological properties, the temperature of the reservoir is an important parameter for any geothermal 460 prospection. The thermal field in the URG has been extensively studied in the past (e.g. Pribnow and Schellschmidt, 2000;Bächler et al., 2003;Baillieux et al., 2013). Accordingly, temperature anomalies are mainly linked to hydrothermal convection zones, which cannot be localized very precisely using classical exploration methods so far. Bär et al. (2021) therefore propose an integrated approach that combines 3D seismic, electromagnetic and gravity data with geothermal gradients from medium-depth boreholes, enabling more accurate mapping of ascending hot brines.

465
In addition to deep EGS projects, underground heat storage will significantly contribute to reduce emissions in the future energy supply. Thereby, seasonal fluctuations of other renewable energy such as solar and wind energy can be compensated.
It is expected that medium depth borehole heat exchangers (MD-BHE) in the crystalline basement have the highest efficiency among comparable technologies and can be applied almost everywhere where the basement is situated near the surface (Welsch et al., 2016). Again, a detailed characterization of the fracture network is essential since open fractures can act as potential 470 fluid conduits that reduce the heat recoverability. In this respect, the presented data can serve as input for thermal-hydraulicmechanical simulations of MD-BHE to be built in the extended URG region.

Conclusions
The fracture network characterization of the Tromm Granite has led to the following conclusions: