Characterization of fractures in potential reservoir rocks for geothermal applications in the Rhine-Ruhr metropolitan area (Germany)

. The importance of research into clean and renewable energy solutions has increased over the last decade. Geothermal energy provision is proven to meet both conditions. Therefore, conceptual models for deep geothermal applications were developed for different ﬁeld sites regarding different local conditions. In Bavaria, Germany, geothermal applications were successfully carried out in carbonate horizons at depths of 4000 to 6000 m . Matrix permeability and thermal conductivity was mainly studied in karstiﬁed carbonates from the Late Jurassic reef facies. Similar to Bavaria, carbonates are located in 5 the east of the Rhenohercynian Massif, in North Rhine-Westphalia (NRW), which quantiﬁcation of the geothermal potential is still lacking. Compared to Bavaria, a supraregional carbonate mountain belt is exposed at the Remscheid-Altena anticline (NRW) from late Devonian and early Carboniferous times. The aim of our study was to examine the potential geothermal reservoir by ﬁeld and laboratory investigations. Therefore, three representative outcrops in Wuppertal, Hagen Hohenlimburg, and Hönnetal were studied. During ﬁeld surveys, 1068 discontinuities at various spatial scales were observed by scanline 10 surveys. These discontinuities were characterized by trace length, true spacing, roughness, aperture, and ﬁlling materials. Joint orientation analysis indicated three dominant strike orientations in NNW − SSE, NW − SE, and NE − SW directions within the target horizon

length was determined according to where µ and c denote mean population frequency and concealed trace length of the recorded discontinuity, respectively. Typically, the concealed trace length is one order of magnitude smaller than the recorded trace length. Note that discontinuities 225 exceeding the height of the outcrop wall could be detected at the nearest level above and/or below the quarry.
The recorded discontinuities were evaluated with MATLAB (2018) after Hudson (2005) and Markovaara-Koivisto and Laine (2012). Markovaara-Koivisto and Laine (2012) developed an open source MATLAB code for the visualization of scanline survey results that was adapted to the purpose of this study. All measurements were recorded according to the metric system of measurement.

Laboratory Measurements: Petrophysical Characterization of Samples
During the field work, loose rock blocks were taken in each quarry for further petrophysical characterization. Compacted limestones, dolomites, and red, dolimitc rocks were named MKB (black Massenkalk), MKY (yellow Massenkalk), and MKR (red Massenkalk), respectively. After sampling the rocks were named according to the sampled quarry, sampled working level, rock type, and sample number. Representative, cylindrical cores with a diameter of 40 mm were extracted from each of 235 these blocks by diamond core drilling perpendicular to the bedding direction. In addition, it was possible for one rock sample (HKW-2-MKB-2-S2) to be cored directly at an exposed rock wall in the outcrop of Hagen Hohenlimburg with a mobile drilling machine. All samples, whether cored in the laboratory or the outcrop, were saw-cut plane-parallel and their end faces were ground square to the maximal possible length l (Table 4). A diameter-to-length ratio of 1:2.5 was aimed for, but this could not always be achieved due to the high density of pre-existing fractures in some blocks. This target sample geometry is generally 240 recommended for triaxial deformation experiments on cylindrical samples (Paterson and Wong, 2005), a topic that exceeds the scope of our current study. All steps of the preparation were conducted with water as coolant and for rinsing removed material.
Following preparation, samples were oven-dried at 60 • C for about 48 h. Basic petrophysical properties were determined on identically prepared samples at ambient conditions, except permeability, which was derived under elevated pressures.
Bulk density ρ geo was calculated from the geometrical volume of the cylindrical samples and their dry masses. Grain density 245 ρ grain was gained from pycnometer measurements on rock powder, produced by crushing and grinding of leftover rock fragments, in compliance with the German DIN 18124 standard. By evaluating the ratio between bulk and grain density, the total porosity was determined according to φ tot = 1 − (ρ geo /ρ grain ). The connected porosity, that is, the externally accessible and connected pore volume, was determined using the difference of the masses of dry and saturated samples with distilled water (see Duda and Renner, 2013).

250
Ultrasonic P-and S-wave velocities, v P and v S , were determined on dry and saturated samples from first-arrival measurements using an ultrasound benchtop unit composed of a waveform generator, two identical broadband ultrasound sensors (1 MHz centre frequency and 0.5 in diameter), and a digital storage oscilloscope (200 MHz sampling frequency). Measurements were performed parallel to the cylinder axis, that is, the drilling direction. Velocities were calculated by dividing the sample length by the determined arrival times less the travel time in the assembly parts. Drained dynamic Young's modulusẼ d and Poisson's ratioν d were calculated by dry P-and S-wave velocities, v P,dry and v S,dry , and bulk density assuming isotropy (see Mavko et al., 2020). Moreover, saturated P-and S-wave velocity, v P,sat and v S,sat , and the density of the fluid-saturated sample was used to calculate undrained Poisson's ratioν ud and Young's modulusẼ ud , assuming isotropy and employing Gassmann's hypothesis (Gassmann, 1951) that S-wave velocity remains unaffected by the presence of a fluid (v S,dry = v S,sat ).
Thermal conductivity λ dry of the dry sample cores was determined by a thermal conductivity scanner using an optical-260 scanning-method at ambient conditions (Popov, 1997). The scanner consists of an emitter and a measuring unit, that is moved along the sample at a fixed distance. The emitted light and heat radiation is focused on the surface of the sample, which heats up the sample pointwise. To ensure absolute absorption of the energy, a part of the sample, usually a strip, was painted black.
Furthermore, it was ensured that the samples for thermal conductivity measurements meet the geometrical requirements for sample size in order to reduce boundary effects. Infrared temperature sensors are located at a fixed distance from the emitter 265 (lead sulfide infrared receiver) and measure the temperature difference of the sample before and after heating. The thermal conductivity was determined by comparison with known standards. In this study, we used the standards of quartz (λ qtz = 1.35 W(m K) −1 ) and titanium alloy (λ Ti = 6.05 W(m K) −1 ). According to the manufacturer (Lippmann and Rauen GbR; TCS No. 2010-013), the determination of thermal conductivity is subject to an absolute error of approximately 3 %.
Permeability k of the samples were determined using Darcy's Law (Darcy, 1856) and a conventional Hoek cell (Hoek and 270 Brown, 1997) to apply axial load, pore-fluid pressure, and confining pressure. Axial pistons and the pre-saturated samples were jacketed by a rubber tube to prevent oil from penetrating the sample, that is, a connection between confining and pore pressure.
In addition, confining and pore-fluid pressure were kept below the axial pressure. A hand pump was operated supplying the axial load (12 MPa). A computer-controlled, high-pressure metering pump was used to apply confining pressure on the sample by compressing distilled water (10 MPa). Axial pore-fluid flow was ensured through central bores in the axial loading pistons.

275
Distilled water was pumped from a water reservoir through the lower end face of the samples either by applying constant flow rates (0.001 to 0.15 l h −1 ) by a second, identical metering pump. The lower axial loading piston was equipped with an outlet pipe discharging fluid pressure to atmosphere (i.e. 1 bar). The temperature and pressure of the pore-fluid was measured to calculate the temperature-and pressure-dependent fluid viscosity according to Wagner (2009)

Results
In the following, the general findings from the field discontinuity investigations, which are found in all outcrops, are presented first, before each outcrop is described individually. Then, the results of the laboratory investigations are introduced.

285
Furthermore, descriptive links between both field and laboratory results are established.

Field Discontinuity Observations
A total of 1068 discontinuity observations were recorded and classified (e.g., discontinuity type, filling, and roughness) by using the field surveys in the three outcrops in Wuppertal, Hagen Hohenlimburg, and Hönnetal (Table 2). It was found that essentially three main sets of discontinuity orientations are predominant in all outcrops (  "Hagen Hohenlimburg". In Hagen Hohenlimburg the scanline surveys were conducted on two exposed rock walls with the orientations NNW−SSE and ENE−WSW. Over a total length of 76.15 m 361 discontinuities were recorded, which were divided into three sets of main orientations ( Table 2). The set 1 lists 134 discontinuity observations, that show a mean common orientation of 224/89 (dipdir/dip) (Table 3) (216), whereas the number of slightly rough (26) and rough fracture surfaces (four) is relatively small. In contrast, on a mesoscopic scale, 55, 155, and 36 discontinuities were found to have smooth, slightly rough, or rough fracture surfaces, respectively. The mean discontinuity trace length was determined to be 3.94 m and lies between the mean trace lengths of the 330 other two outcrops. The mean discontinuity distance is similar to the one determined during the surveyed discontinuities in Hagen Hohenlimburg and was determined to be 0.27 m.

Laboratory Characterizations
Samples could be obtained from all outcrops of dolomite and limestone, although it was challenging to obtain samples from the Hönnetal due to the lack of revealed outcrops. Most of the exposed Hönnetal limestone rocks were pure and showed a very 335 low grade of dolomitization. Nevertheless, some dolomites have been sampled. Dolomite samples from Hagen Hohenlimburg occurred as highly fractured rocks with karst and highly porous formations ranging from a few millimeters to several centimeters. In total, from the collected rock blocks of the Wuppertal, Hagen Hohenlimburg, and Hönnetal outcrops six, eight, and six samples could be prepared, respectively. Furthermore, it was possible to drill a sample directly from the existing rock face in Hagen Hohenlimburg, that is described separately below.
oriented perpendicular to this stress tend to be closed (Lorenz et al., 1991). It can therefore be assumed that higher fracture permeability can be expected in the N−S direction.

Filling and Surface Roughness of the Fracture Network
Information about fracture fillings and surface roughness can be used to estimate the reservoir behaviour under in-situ conditions, since fillings and roughness have a direct influence on the elastic, hydraulic and thermal properties of the reservoir.  It is very complex to determine fluid flow by both numerical and analytical methods along two irregular faces, that is, natural fracture surface. However, the impact on fluid flow along rough fracture surfaces was already shown and discussed (e.g., Brown, 1987). In the scope of this study, fracture roughness was ascertained qualitatively regarding the applied method. The results indicate predominating smooth fracture surfaces on the field scale (> 10 -1 m) and mainly slightly rough fracture surfaces on the mesoscopic scale ( 10 -1 m). In the literature roughness on field scale are typically described as waviness or straightness 505 (ISRM, 1978). Verifying the determined roughness is only possible with difficulty. However, considering the numerous tectonic events and its repetitive reactivation of pre-existing fractures, smooth fractures tend to be reasonable.

Connectivity of the Fracture Network
The results of our scanline suveys revealed that discontinuities with short trace lengths occur predominantly on the field scale. This observation was made in all three quarries. Almost 70 % of the recorded discontinuities classified as "both ends 510 visible" tended to be between 1 and 2 m long. It is conceivable that the observed trace lengths increase with increasing outcrop length and height. The recorded end types suggest that 40 % of all discontinuities investigated exceeded the observed outcrop height. In other words: 40 % of all measured trace lengths might be underestimated by the method. This source of error was attempted to be minimized by the censored semi-trace length analysis according to Priest and Hudson (1981). Contrary, 60 % of all joints counted showed a start and end tip (Table 2). However, a full 3D discontinuity analysis of the entire quarry 515 will likely reflect better trace length results depending on the accuracy. At this point, economic considerations must be made regarding the required survey accuracy and associated costs. Regarding our joint analysis from Wuppertal, approx. 70 % of surfaces are more likely to occur in the reservoir. Discontinuities sets with short trace lengths occurred predominantly in all three outcrops. Almost 70 % of the recorded discontinuities classified as "both ends visible" tended to be between 1 and 2 m long. We compared the mean trace lengths with the mean true discontinuity spacing in order to get an idea of a coherent fracture network in the reservoir. Taking all measurements into account, we expect comparable probabilities for interconnecting fracture networks in the deep reservoir. However, based on the observations of many karst formations and altered host rocks by 590 hydrothermal veins in Hagen Hohenlimburg, we conclude that the permeability is greater in the regions of the Devonian reef facies than in the other areas investigated. This assumption was also confirmed by our laboratory measurements, whose results suggest the presence of heterogeneities that may favours fluid flow.