Hydrogeochemical modelling of groundwater in a fractured Carboniferous sandstone aquifer

Deep groundwater monitoring in tight bedrock is not commonly performed and rarely documented. For a case study in Bochum, NRW, deep monitoring wells allow groundwater investigation in the outcropping Carboniferous down to a depth of 186 m. Groundwater monitoring was carried out by combining sampling of the monitoring wells and local springs at the site. Inverse models were developed using the hydrogeochemical code PHREEQC to understand the natural geochemical processes of deep groundwater in the coal-bearing sandstone. The identified hydrogeochemical processes are pyrite oxidation in shallow fracture systems and Na-Ca/Mg cation exchange in deep fracture systems. Core samples reveal fractures containing Fe-dolomite and calcite cements. A freshening process dominates the deep fractured rocks, where fresh water is flushing a saline aquifer. The data presented here support the planning of further monitoring wells and help to improve future modelling approaches.


Introduction
The German Ruhr District in North Rhine-Westphalia (NRW) has been a region of intense hard coal mining since the 18th century.The shutdown of the Ruhr coal mining activities in 2018 marked the transition to a post-mining era (Hager 2018;Drobniewski 2018;Melchers and Goerke-Mallet 2018).Currently, the Ruhr region is in the process of a structural transformation.Hard coal as an energy source is to be progressively replaced by new regenerative forms of energy in NRW and throughout Germany (MWIDE NRW 2021; Moeck et al. 2021;BET 2013;BMWi 2010;UBA 2015).Shallow and deep geothermal energy benefit from this development.The advantages of geothermal energy are its base load capability and low space requirements, even in limited urban spaces.Furthermore, the subsurface offers a high heat storage potential (Bracke and Huenges 2022;Born et al. 2022;LANUV NRW 2015).Innovative concepts for low-enthalpy energy generation or subsurface energy storage in the Ruhr District are being proposed and evaluated (LANUV NRW 2018;Niemann and Schreiber 2018) and some are even in the pilot stage (Hahn et al. 2022;Bussmann et al. 2023).
In this context, underground research laboratories are important as analogue studies since they help to investigate the basic physical-chemical-biological understanding of sites with similar geological characteristics (Bracke and Huenges 2022).Instead of performing experiments on a large industrial scale, it is more practical to run subsurface field experiments on a mesoscale, specifically in the 10 to 100 m range.This approach considers the natural anisotropy of the rock, which is not the case in laboratory-scale experiments using rock core samples.For this reason, Fraunhofer IEG, the Bochum University of Applied Sciences and the Ruhr-University Bochum established a 10,000 m 2 in situ laboratory within the 50 km 2 geothermal permission field "Zukunftsenergien" for reservoir, advanced drilling and geophysical experiments under real case conditions (Bussmann et al. 2015;Bracke et al. 2017).
The site in Bochum was initially chosen to install and explore the Hot Dry Rock (HDR) technology for sandstones down to 4000 m depth (Rubitec 2003) to be able to supply urban district heating with geothermal energy.Today, research is focused on applied energy concepts and geothermal technologies in naturally fractured reservoirs as well as in mine voids.So far, two deviated borehole heat exchanger (BHE) arrays with heating and cooling capabilities, a mine thermal energy pilot storage (MTES) system and a set of deep groundwater monitoring wells associated with research boreholes have been installed (Bussmann et al. 2015;Bracke et al. 2017;Hahn et al. 2019Hahn et al. , 2022)).
We present a synthesis of results from a well drilling and coring campaign combined with groundwater and spring monitoring for this specific site at Bochum.We aimed to understand rock-fluid interactions to optimise future in situ circulation and storage experiments (Erstling et al. 2019;Hahn et al. 2022;Seidel et al. 2022;Nehler et al. 2022) and to enable a future distinction between natural groundwater variations from experiment-induced ones (baseline monitoring).We show that a freshening process dominates the deep fractured rocks, where fresh water is flushing a saline aquifer.With the data presented here, we support the planning of further monitoring wells and help to improve future modelling approaches.

Carboniferous subsurface geology of the Ruhr area
The Ruhr Coal Basin is part of an external fold and thrust belt of the Variscan orogeny in Europe.The fold belt stretches from Ireland and southern Britain through France, Belgium and Germany to Poland (Drozdzewski 1993).The overall subsurface geology and the tectonic and stratigraphic framework of the upper 1500 m of the Carboniferous strata are well-known based on mining activity.In the southern Ruhr river valley, the coal mining era began at the beginning of the 18th century, mainly focusing on shallow hard-coal seams, and the mining activities continued into deeper collieries to the north (Huske 2006).
Sediments and sedimentary rocks in this region represent a more than 5500 m thick, mostly molasse-type sequence of Late Carboniferous age and are mainly composed of siliciclastic rocks with about 250 intercalated coal seams (Drozdzewski 1993;Brix et al. 1988).Sediment deposition and sedimentary rock deformation took place penecontemporaneously and resulted in the Variscan foredeep and foreland basin, which delimits the present-day Ruhr basin (Oberste-Brink 1948).The basin strikes in a southwestnortheast direction and has a length of 150 km and a width of 80 km across the strike.The structural framework of the strata in the southern Ruhr District is characterised by steep dip angles and folds (Drozdzewski 1993).
During the Variscan orogeny and post-orogenic sedimentation, the depositional environment gradually changed from open to shallow marine and coastal plain and deltaic settings (Oberste-Brink 1948;Drozdzewski 1995).Coal formation of the Ruhr basin commenced in the Namurian C (ca 317-316.5 Ma) and reached its maximum during the subsequent Westphalian (316.5-305Ma) (Deutsche Stratigraphische Kommission (DSK) 2016).The characteristic claystone-siltstone-sandstone strata and coal seams are uniformly distributed across the Ruhr basin, but some facies may thin out or are regionally absent in places.Wanless and Weller (1932)  transgressive marine to continental (including coal beds) depositional sequences.These authors coined the term 'cyclothem', now widely applied in the literature for cyclical sedimentary successions in the Carboniferous.Later, several authors redefined the term cyclothem sensu Wanless and Weller (1932) and suggested the label 'continental cyclothem' for repetitive successions of (continental) coarseto fine-grained sedimentary rocks and coal seams.This adopted terminology was frequently criticised for misusing the original concepts (Heckel 2008).We acknowledge this criticism but apply the terminology as it is rather well established in the regional literature (Stehn 1988;Drozdzewski 1993Drozdzewski , 1995;;Wrede 2012).

Bochum investigation site
The study site (7°16 0 27 00 E / 51°26 0 47 00 N) is located in the southeastern area of the municipality of Bochum (Fig. 1).Small valleys to the west and east, each feeding a creek via ephemeral springs, delimit the study area (Fig. 1).The Upper Carboniferous strata are covered by Quaternary loess sediments and weathered Carboniferous bedrocks, forming a regolith of about two metres in thickness.The strata underlying the weathering zone form a fractured rock aquifer with low to moderate permeability.The study site is within a syncline built by Westphalian sandstones (Fig. 2: Bochum Formation; 313.5-315Ma, Deutsche Stratigraphische Kommission (DSK) 2016).According to Stehn (1988), Carboniferous groundwater flow occurs mainly in fracture systems related to fault zones and folded strata.Therefore, the overall structural setting merits attention.
In 2015, three deep groundwater-monitoring wells were drilled into the fractured sandstone.Two monitoring wells are located in the same stratigraphic interval and were sampled in a 2-week interval.The targeted sandstone succession was first encountered through core drilling in research well R1 at approximately 121-132.5 m b.g.l. and belongs to the Fig. 2 Schematic stratigraphic column of the Bochum study site according to Drozdzewski (1993), Fiebig (1957) and Stehn (1988).The average thickness is about 250 m.Sandstones shown here are marker horizons Abb. 2 Schematische stratigraphische Kolonne des Bochumer Untersuchungsgebietes nach Drozdzewski (1993), Fiebig (1957) und Stehn (1988).Die durchschnittliche Mächtigkeit beträgt etwa 250 m.Die hier gezeigten Sandsteine sind Leithorizonte Schöttelchen unit.Lithostratigraphically, the Schöttelchen unit is divided by three thin coal seams, interbedded with a fine-to coarse-grained sandstone (Fiebig 1957;compare Figs. 2 and 4).
A limited number of fault structures are documented on regional geological maps (Bezirkregierung Arnsberg, Abt. 6 (BRA) 2018; Stehn 1988).Two major fracture sets were identified in boreholes.Borehole imaging by Jagert (2016) shows bedding with a local average northwest-oriented dip direction, typical for the Variscan orientation of strata, with dip angles of about 65-70°(Fig.3a, b).The fracture orientations were interpreted as conjugated shear joint sets.This type of fracture system was mapped in all borehole images and is the dominant fracture system (Jagert 2016).

Mineralogy and diagenesis of Carboniferous Ruhr sandstones
Previous regional petrographic studies describe silty claystone and hard coal layers (e.g.Wegehaupt 1962;Esch 1962;Menyesch 1978).The diagenesis of the regional Carboniferous sandstones was studied in detail by Füchtbauer (1974Füchtbauer ( , 1979) ) and Wegehaupt (1962).Cement phases of the shallow burial domain are mainly kaolinite and siderite cement.These are overgrown by quartz and ankerite cement.The accessory mica is mostly muscovite, and less commonly, biotite.Chloritisation of biotite continued during much of the burial history.

K
Generally, the maturity of the Upper Carboniferous sandstones is comparably low.Quartz, feldspars and detrital mica (biotite, muscovite), combined with illite and chlorite, dominate these rocks (Hucke 2002 and references therein).Accessory phases include dolomite-ankerite and kaolinitedickite cements.Pyrite, which is mostly associated with fractured sandstones and calcite cement occluding fractures, has also been reported.
The facies and grain size of the various lithologies and their diagenetic properties control the nature and type of fracturing.Coarse and fine sandstone units display the highest rock strength, according to Hucke (2002).Quartz cement significantly contributes to the overall mechanical properties of most lithologies (Hucke 2002).

Cored samples: mineralogy, porosity and permeability
In the context of this study, cores from well R1 with a diameter of 85 mm were analysed.Analyses were performed to compile data on the specific mineralogical spectrum of the underlying strata at the study site.Mineralogy, in turn, is relevant for the understanding of fluid-rock interaction.Samples described here relate to the Schöttelchen sandstone of the Lower Bochum strata (Jagert 2016).The PVC-screen intervals of groundwater monitoring wells O3 and O4 were located in this specific sandstone section to enable sampling of transmissive fractures (Fig. 4).
Thin sections of Schöttelchen sandstone were fabricated from samples taken at coring depths of 121, 122 and 126 m.Some samples included an open fracture.The initial description was made using polarised light microscopy followed by X-ray analyses.
Scanning electron microscopy (SEM) images of backscattered electrons (BSE), as well as energy dispersive X-ray analysis (EDX), were performed at the Microanalytical Lab of the Faculty of Geosciences, Ruhr-University Bochum (dataset from Weisenberger et al. 2020).Scans were generated using a Zeiss Merlin Gemini 2 field emission scanning electron microscope with a maximum spatial resolution of 0.8 nm.The actual resolution of an SEM analysis depends on several factors, such as applied voltage (kV) and material mixture and will not reach 0.8 nm.BSE images on thin slides provide information on the mean atomic mass by means of grayscale images and hence provide information about its chemical composition.EDX is a standard method for the identification as well as quantification of  Stehn (1988) the elemental composition of samples via X-ray spectral analysis.
Additional analyses such as X-ray powder diffraction analyses (XRD), mercury porosimetry and gas permeability measurements were performed at the GFZ German Research Centre for Geosciences, Potsdam (Reinsch et al. 2019).XRD analyses of the Schöttelchen sandstone were obtained with a Malvern Panalytical Empyrean X-ray diffractometer (used software: Panalytical High Score Plus 3.0.5;database: ICSD; 2-Theta).Porosity determinations were performed using a 2000 WS Porosimeter (Carlo Erba K Instruments).The permeability of the Schöttelchen sandstone was quantified using a conventional gas permeameter with argon as a gas medium.The Klinkenberg correction method was applied to calculate the intrinsic permeability of the rock samples, following the approach of Tanikawa and Shimamoto (2009).

Water sampling
Data on the wells and analytical results are shown in Table 1.Groundwater sampling data from the monitoring wells were combined with sampling data from two adjacent spring valleys (Königsbüscher Wäldchen and Kalwes valley) (Einarsson et al. 2020).The filter screen of the two sampled observation wells O3 and O4 (Figs. 1 and 4; Table 1) were installed at different depths: The transmissive section (gravel pack) of well O3 is between 84.2-107 m and that of well O4 between 161-186 m below ground level.Before sampling the wells, pumping was performed until stable specific electrical conductivities, and stable redox potentials (Eh) were achieved.During sampling, the following in situ parameters were determined: Temperature, pH, dissolved oxygen, specific electrical conductivity and redox potential.HCO3 -was measured by titration in the laboratory immediately after sampling.Water samples were filtered (0.45 μm pore size), acidified (HNO3) and cooled for subsequent analysis of cations and anions.Anions were analysed using Metrohm ECO IC ion chromatography.NO3 -was measured using non-acidified samples.Cations were measured with an ICP-OES (Optima 8300) instrument from Perkin Elmer at the Fraunhofer IEG, Bochum.

PHREEQC modelling
The hydrogeochemical software PHREEQC Version 3.6.2.15100 was used to simulate the hydrogeochemistry of groundwater from the sites (Parkhurst and Appelo 2013).The program's capability includes choice and extension of thermodynamic databases, hydrogeochemical equilibrium calculations, and calculation of elemental concentrations, species distribution, pH, redox states, saturation indices and mol transfers.
Solution modelling within PHREEQC was performed using characteristic water chemistry data from a set of 202 water samples of shallow groundwater (well O3), deep groundwater (well O4), and two spring waters described above.First, hydrogeochemical data were specified in PHREEQC, and results were analysed for charge balance and mineral saturation indices (database: PHREEQC.dat) with the extended mineral phase ankerite, according to thermodynamic data from Hellevang et al. (2013).Inverse model simulations of the reaction pathways between the initial (shallow groundwater) and final solution (deep groundwater) included mineral precipitation, dissolution and ion exchange.A similar approach for mine water reservoirs has also been illustrated by Perry (2001).Natural organic matter CH2O (represented as carbohydrate, according to Appelo and Postma 2005) was integrated as a phase to simulate carbon influx into the system.

Aquifer mineralogy of the Schöttelchen Sandstone
Thin-section analysis of sandstone samples under a polarising microscope yielded the following mineralogical results (given in percentage area) for the rock matrix (Fig. 5).The sandstone comprises about 30% quartz, with 2-3% feldspar in the form of albite.Lamellar twinning, according to the albite law of plagioclase, is observed.Muscovite as a detrital mineral is frequently observed and amounts to about 5-8%.Biotite, another detrital mineral, is locally present but represents less than 1% of the host sediment.Furthermore, microcrystalline quartz cement occludes pore space and replaces matrix sediments (3-5%).Clay and quartz cement are the main cement types identified in the samples.As indicated by the presence of partially degraded feldspars, some of the clay was derived from the diagenetic alteration of feldspar and other metastable minerals.Hard coal in the form of opaque matter reaches about 5-12%.
Thin sections of samples taken between 122 and 126 m depths were studied by BSE and EDX.The outcome is consistent with the data reported above.Figure 5(g-h) depicts one of the characteristic open fractures and their diagenetic infill.In the original core sample, this fracture shows a clear opening width and is part of a conductive zone that was already noticed during drilling activities due to watercirculation losses.Weisenberger et al. (2020) and Olschowsky (2020) report the following additional components based on EDX spectra: plagioclase (albite), muscovite, pyrite, hard coal, and various Al-alteration products (clays).Ferroan dolomite occludes pore space and portions of the still open fractures.In fracture-filling carbonate cements, the predominant cations (Fe 2+ , Ca 2+ and Mg 2+ ) suggest a complex solidsolution series between dolomite and ankerite.For the sake of simplicity, we here apply the term Fe-dolomite, and we acknowledge the significance of this phase in the context of PHREEQC-modelling.The Fe-dolomite exhibits zoning that is visible (Fig. 5).EDX measurements reveal that the Mg 2+ content decreases towards the centre of the fracture, giving way to ankerite.In essence, the mineralogy of these cement phases is not trivial but detailed research focussing on the different paragenetic phases in these rocks is beyond the aims of this paper.
Porosity data based on mercury porosimetry point to a very low porosity for the studied samples.These are 1.4-2.8%,with bulk densities between 2.6-2.7 g/cm 3 (Reinsch et al. 2019).The impression gained is that of a low-permeability sandstone with a pervasively cemented matrix.In the case of a permeametry analysis for one Schöttelchen sandstone sample from 126 m, an intrinsic matrix permeability of k = 10 -19 m 2 was calculated.No flow could be determined for further subsamples of this sample, even if the inlet pressure was increased to 20 bar.
Following sampling in groundwater well O4, the drawdown was recorded and evaluated according to the Theis type curve.The drawdown reached about 21.

Hydrogeochemical classification and mineral saturations
The water data selected for modelling display a charge balance error well below 3.3% (Table 1).Moreover, concerning the spring samples, analyses selected for this study were sampled after rainfall events in September and tended to correspond to groundwater discharge.Classification of water types was made according to Langguth and Voigt (2004): The percentages of the equivalent concentrations (meq/l) of the ions to their cation and anion sums are used to classify the water type.Ions with a fraction of 20% and more are named according to the order of their fractions, with cations taking first place over anions.Ions with proportions of at least 50% and more are highlighted (in italics).
The shallow groundwater from monitoring well O3 has a hydrogeochemistry typical for recharge areas of this region: The hydrogeochemical composition points to a Ca-Mg-HCO3 type water.This type of water corresponds to recharge water, which is saturated with atmospheric CO2 resulting in the dissolution of carbonate minerals (e.g., cal-cite and Fe-dolomite cement) in the aquifer, leading to an increase in dissolved Ca 2+ , Mg 2+ and HCO3 -.
Due to its composition, the deep groundwater of well O4 is classified as typical ion exchange water of the Na-HCO3 type.The concentration of Ca 2+ and Mg 2+ is significantly reduced compared to the samples from monitoring well O3.The same pattern is found in the case of SO4 2-.In contrast, HCO3 -, and in particular Na + , are noticeably enriched.Specific electrical conductivity increases with depth from 482 µS/cm (O3) to 707 µS/cm (O4).The pH increases from 7.1 (O3) to the alkaline range of 8.2 (O4), comparable to seawater.
The spring sampled to the east of the study area (Fig. 1) is located in the nature reserve Königsbüscher Wäldchen, and we find limited evidence for anthropogenic activity.The composition of the groundwater discharging from gullies corresponds to the Ca-Mg-HCO3 water type.The pH is neutral at 7.7, but can shift to alkaline values during the season.The specific electrical conductivity measured on September 15, 2020, was 644 µS/cm.
The sampled spring in the western Kalwes Valley (Fig. 1) is affected by anthropogenic activity, as rainwater retention basins have been constructed that drain into the valley.The composition of the ions corresponds to a Na-Ca-Cl-HCO3 water type.The genesis of this water type is not necessarily obvious but argues for the mixing of different waters.The pH measured on September 1, 2020, was 7.9, but seasonal variations can reflect even more alkaline conditions.The specific electrical conductivity of the water samples varies strongly, which is considered further evidence for mixing surface runoff and groundwater discharge (Appendix, Fig. A1).
The Schöttelchen sandstone is cut by both spring valleys and is considered an associated sandstone aquifer, besides other sandstones and fractured siltstones (Fig. 3a, b).Seidel et al. (2022) have published local groundwater modelling that yields a rough velocity estimate of about 30 m/yr for shallow groundwater (Contour map in Appendix, Fig. A2, porosity: 6%, hydraulic conductivity: 5.0E-06 m/s).Actual velocities in deep bedrock are expected to be even slower.The manually logged groundwater level (delta) varies by 1.90 m (well O3) and 3.10 m (well O4) between fall (lowest level) and spring (highest level).
When considering the S.I., it is generally assumed that values near zero represent an equilibrium state (Table 1).The deep groundwater in monitoring well O4 is in equilibrium with calcite and dolomite.Very low iron concentrations were measured for the two spring samples and well O4.Nevertheless, iron is required for modelling mineral phases such as ankerite, pyrite, siderite or amorphous ferrihydrite.

Conceptual model for deep Na-HCO 3 groundwater origin
The groundwater samples taken from well O3 (Fig. 4) are best defined as shallow groundwater, affected by groundwater recharge and dominated by dissolution of carbonates (calcite, Fe-dolomite) in the host aquifer.When evaluated by its saturation index and Fe 2+ and SO4 2-concentrations, a low degree of pyrite oxidation is clear.The aqueous fluids change with depth due to water-rock interaction, including ion exchange and/or dissolution/precipitation reactions (well O4).Concerning the deep groundwater, a HCO3 - source, albeit one that is not well constrained, is found.In addition, the hydrogeochemistry is significantly depleted in SO4 2-, indicative of sulfate reduction and/or pyrite formation.In addition, an unknown chloride source affects the water geochemistry.
Hydrogeochemically different groundwaters could also be transported by regional flow fields, which mix with the aquifer on site.This is not expected because the slow groundwater flow velocity of about 30 m/yr and the low hydraulic conductivity of the Carboniferous at about 10 -6 -10 -9 m/s (Seidel et al. 2022) suggest high dispersion of transported solutes.
From the above-described processes, the following mineral phases are implied that must be integrated into the model: 1) Calcite as well as the two endmembers of the solid solution between dolomite and ankerite.Although not observed in BSE data so far, Fe-dolomite generally coexists with Fe-calcite or siderite.For this reason, siderite was also integrated.2) Pyrite as a source or sink for S and Fe and 3) besides carbonates, CH2O is an organic source for carbon.4) Halite (NaCl) provides an unspecified chloride source, and 5) ion-exchange sites for Na + , Ca 2+ , Mg 2+ and K + , given that the very significant enrichment in Na + in the deep groundwater cannot be explained otherwise.A cation exchange capacity (CEC) with 44 mmol(eq)/kg for Ruhr sandstone was determined by Paas (1997) and is, in the context of the PHREEQC model, equilibrated with the groundwater of well O4 to evaluate major fractions of adsorbed cations: Since in situ fracture porosities are not available so far, a geometrical recalculation was made between mmol(eq)/kg and mmol(eq)/l.Appelo and Postma (2005, p. 248) provide a way to calculate the water-rock contact.However, this way is valid for loose aquifer sediment and not for fractured bodies, so another way must be found: Assuming a 1 × 1 m fracture plane with an aperture of 0.8 mm (Olschowsky 2020) gives 0.8 litres per fracture.This fracture has two planes (sides), which results in an area of 2 m 2 .2m 2 divided by 0.8 litres result in 2.5 m 2 /l groundwater.Following Cvetkovic (2010), a fracture "rim zone" is defined here in exchange with the groundwater.Assum-K ing a thickness of 1 mm, the mass of the previously defined fracture is: 2.5 m 2 /l times 0.001 m "rim zone" = 0.0025 m 3 "rim zone" volume per litre.This volume times an average density of 2650 kg/m 3 of Ruhr sandstone results in 6.63 kg of fracture material ("rim").The 44 mmol(eq)/kg, according to Paas (1997), multiplied by the "rim" mass, results in 292 mmol CEC per litre of groundwater with fracture contact (meq/l CEC).This rough derivation is simplified and does not include complex fractures, porosity and diffusion within the rim zone.
Dissolution and precipitation reactions for quartz, albite, muscovite, biotite, and kaolinite were not considered in the model due to the fact that aluminium concentrations in the water samples are very low and below the detection limit of our ICP-OES facility.We suggest that these low values point to insignificant alteration/mass transfer of aluminosilicates between rock and water in the aquifer.

PHREEQC inverse modelling results
Based on PHREEQC modelling, 10 inverse model variants result from the setup described above.Each of these has the potential to explain the Na-HCO3 water chemistry (Table 2).In the context of modelling, some concentration input parameters were adapted to allow for their computing.An error tolerance was set at 3.5% (uncertainty) to provide robust results.Phase mole transfers suggest that Na + exchange is a dominant process in groundwater.A total of 5.2 mmol/l Na + are released into the solution exchanging with Ca 2+ and Mg 2+ .The programme did not require K + for any of the variants as an exchange reaction.
In particular, most model outcomes suggest the preferential dissolution of ankerite or dolomite and, in some cases, calcite in an undersaturated phreatic environment.The same applies to siderite (FeCO3).All models suggest that pyrite (FeS2) is precipitated at 23 mg/l.Dissolution of carbohydrate (CH2O; 20.2 mg/l) is suggested by all model variants.Carbohydrate in the studied system may have a microbial, or more generally, organic matter origin (e.g., coal or other organic sediments).Halite (NaCl) is easily dissolved (59.73 mg/l) and provides an additional source for Na + and Cl -.Specifically, halite dissolution is a prerequisite; otherwise, the mass balance of Cl -is not achieved.The source of halite in the groundwater and the host aquifer, respectively, is not well constrained.
The results of cation-exchange equilibration in the applied model point to an equivalent fraction of approximately 49.3% for Ca 2+ , 30.5% for Mg 2+ , 16.6% for Na + and 3.7% for K + adsorbed on exchange sites while in equilibrium with the groundwater of well O4 (Table 3).Minor fractions of the exchanger are loaded with Fe 2+ and neglected in the applied model.The maximum transferred fractions  do not reach 100%, so we assume that the aquifer is still loaded with Na + (Table 3, column 6).Ion exchange could therefore continue with subsequent groundwater flow.The model considers only estimated CEC and does not include the influence of pH changes.There is a need for further research on the actual CEC of an entire fractured aquifer measured by field methods.

Interpretation and discussion
The Carboniferous sandstone studied here represents an example of a typical sandstone aquifer in the southern Ruhr District, characterised by a low porosity and low matrix permeability.Tectonically controlled, systematic fracture networks act as pathways for the actual fluid flow.The storage properties of this bedrock refer, on the one hand, to a fracture porosity and, on the other hand, to a matrix porosity.
Pumping tests determine the fracture porosity as it is part of a mobile domain.The low matrix porosity, an immobile domain, is challenging to access, probably only by lab experiments.
The mercury intrusion porosimetry and permeametry analyses are largely in agreement with previous work on the site (Hartung 2017: k= 10 -19 -10 -16 m 2 ; ntotal = 0-5.5%).It should be noted that the mercury intrusion method measures the connected porosity, as the mercury does not reach the closed pores.
The shallow groundwater is characterised as a Ca-Mg-HCO3 water type, while the deeper groundwater is a Na-HCO3 type.Both compositions remain almost constant over the time of sampling.In a small valley in the west of the study area, surface processes seem to dominate.The Na-Ca-Cl-HCO3 water type represents a mixed water type.This becomes evident from the dynamics in the measured values over the sampled period (Appendix, Fig. A1).The small valley in the eastern parts of the study area is a nature reserve, the water type shows only little dynamics and is identified as Ca-Mg-HCO3 type.
The PHREEQC model considers readily soluble minerals.A dissolution of either calcite, dolomite, ankerite, siderite or a combination of these minerals was identified.Ankerite and siderite, besides pyrite/S 2-, may act as a source or sink for Fe 2+ .Carbonate dissolution is a key part for the groundwater processes and leads to the uptake of HCO3 -.
The shallow groundwater from well O3 shows elevated concentrations of SO4 2-(38.8mg/l) and Fe 2+ (0.9 mg/l).Elevated SO4 2-and Fe 2+ may be explained by dissolution of pyrite, which is frequently contained in the Carboniferous sediments and the hard coal.However, the data are not strong enough to provide a specific explanation.The oxygen required for this process would probably be provided by rainwater from the vadose zone and transferred to the groundwater by mixing.Additional SO4 2-from agricultural sources is unlikely since no NO3 -, commonly considered as indicator for agriculture input, was measured.However, nitrate reduction via pyrite is a conceivable process, too, so an anthropogenic influence cannot be excluded in its entirety (Wisotzky 2015).
The groundwater of the deep well O4 is nearly sulfatefree (3 mg/l).The inverse model points to pyrite (FeS2) precipitation from groundwater.It should be stated, that no sorption or anion exchange was integrated into the modelling, which could explain an alternative loss of SO4 2-and Fe 2+ .The pH of groundwater affects the ability of minerals to adsorb cations or anions.The neutral point at a given pH where protons compensate all other charges is called the "point of zero charge" (PZC) and is mineral specific (Appelo and Postma 2005;Adair et al. 2001).In general, in groundwaters with high pH, which is the case here, cation exchange takes place.
Na + exchange is a dominant process in the inverse model.Upon contact with the shallow groundwater, a majority of the Na + is released from the aquifer, and Ca 2+ and Mg 2+ are bound in its place (Eq. 1, according to Appelo and Postma 2005): Genetically, a Na-HCO3 water type points to what is called a freshening process, a feature well described in the literature (Cederstrom 1946;Foster 1950;Schwille 1955;Walraevens and Camp 2004;Appelo and Postma 2005).Schwille (1955) noted that these exchange waters commonly originate in low permeability rocks or wells with low productivity.Otherwise, the exchange sites would not K regenerate at high flow rates and become depleted.This line of evidence is in agreement with the low-permeability rocks studied here.
According to Walraevens and Camp (2004), the freshening trend takes place when freshwater is in contact with saltwater, resulting in the formation of a transition zone (mixing zone).In this mixing zone, the depletion of the Cl -concentration due to mixing (dilution), and excess of Na + and loss of Ca 2+ , occurs due to cation exchange.Note that the groundwater of well O4 is also strongly depleted in Mg 2+ .In contrast, no significant cation exchange could be detected for K + .
It is noticeable that HCO3 -is elevated for the deep groundwater.Appelo and Postma (2005) state that this is related to ion exchange: When Ca 2+ is exchanged for Na + , the water becomes "soft", i.e. undersaturated for calcite/ dolomite and dissolves even more carbonates.For both minerals, saturation is coupled with ion exchange and in equilibrium (compare Table 1).The pH of the deep Na-HCO3 water increases to > 8 due to dissolution of calcite/ dolomite.
Besides carbonate dissolution and ion exchange, there is another reason for increased HCO3 -concentrations: as Cederstrom (1946) noted in his study, low SO4 2-values are associated with contemporaneously increased HCO3 - and most likely a reduction in combination with organic material has taken place.Under anoxic conditions, SO4 2- is transformed into S 2-and HCO3 -where organic matter is degraded.The weak H2S odour points in this direction, as well as the actually elevated HCO3 -concentration in well O4 (Eq.2, according to Wisotzky (2015) with organic matter as basic notation CH2O): In all inverse model variants, halite was added to control for an unknown Cl -source.Chlorine may well be present in hard coal and minerals such as mica, feldspars or siderite (Yudovich and Ketris 2005).The Cl -source may also be anthropogenic (e.g., road salt), but this mechanism does not apply to deep wells.Finally, another possible origin may be deep saltwater.Along these lines of evidence, formation waters of the Ruhr District are known to become more saline with increasing depth and qualify as brines similar to many sedimentary basins worldwide (Michel 1963;Wedewardt 1995 and references therein).This pattern correlates with the Na-HCO3 character of the local deep groundwater, indicative of a freshwater-saltwater mixing zone.

Drawing a regional picture
In the framework of this fractured aquifer study, a larger regional picture can be drawn: Intensive coal mining was carried out in the region.At the site, mining was carried out locally near the surface to a depth of 65 m in the 1950s.The monitoring wells are located outside these zones, so that affected groundwater is unlikely.About 800 m to the south, the Klosterbusch colliery was in operation until 1961 (Huske 2006).The recorded depth of mining was 564 m, so that the natural groundwater chemistry may have been affected by the mine dewatering and the subsequent flooding at that time.These effects on the hydrochemistry may persist to the present day.
The study area is located between two dewatering areas of the Ruhrkohle AG (RAG), which are currently still in operation: Robert Müser (4.8 km distance) pumps 10.6 million m 3 of mine water per year.Friedlicher Nachbar (7 km distance) pumps 8.5 million m 3 of mine water per year (LANUV 2018).
Both stations produce brackish water, recognizable by the specific electrical conductivity: Friedlicher Nachbar at 2890 µS/cm from a depth of 265 m, and Robert Müser at 6150 µS/cm from a depth of 554 m (Wedewardt 1995).These dewatering stations are part of a larger drainage system of flooded, abandoned galleries in the Ruhr District (Drobniewski 2018).The operation is intended to protect shallow groundwater from rising mine water.Hydrogeochemical changes to very distant fractured aquifers are possible.If dewatering occurs, freshwater may percolate deeper and flush saltwater-saturated rock blocks.Thus, this kind of anthropogenic influence could provoke ion exchange and may also explain the unknown chloride source.These assumptions need to be confirmed by further field measurements and numerical modelling of regional deep groundwater flow fields.

Conclusions for future in situ experiments
Based on the mineralogical characteristics of the aquifer material, the hydrochemical monitoring data and the groundwater dynamics for the study area, the following conclusion can be stated regarding the establishment of an in situ test site for geothermal applications: Concerning hydrogeochemical modelling, ion exchange has been significant.This implies that future numerical models should consider not only the main components of Carboniferous sandstones (quartz, feldspar, mica, clays) but also processes in fractures that are different from that in the host rock.According to Appelo and Postma (2005), a "chromatogram" should be recognizable in the groundwater chemistry, i.e., in addition to the Ca-Mg-HCO3 type and the Na-HCO3 type, a K-HCO3 and an Mg-HCO3 type should be found in the aquifer layers.This should be verified with additional groundwater monitoring wells.Such considerations of the region's water quality have not been provided to date and should be supplemented with regional studies.The studied area includes a watershed: If groundwater tracer tests are planned, the tracer could be split into two directions.This must be taken into account in future field tests.The fracture system in local wells appears to be very native, i.e., more valuable for reservoir considerations for geothermal research, unlike quarry samples.This concerns the porosity, permeability, and composition of fracture veins.However, surrounding mining may have had an impact on fracture mineralogy through drainage, percolation and the associated alteration.
This exploratory-descriptive study documents baseline monitoring so that the impacts of future subsurface experiments are detectable.This study's data can now help improve reactive transport modelling associated with field tests.Nevertheless, the study still needs to be expanded more regionally.The actual groundwater stratification, flow velocity and effective porosity must be determined in the field to improve documented results.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.(Seidel et al. 2022).Background of the map is a shaded relief from digital terrain data (Geobasis NRW) Abb.A2 Grundwassergleichenkarte des mit SPRING © modellierten Gebietes (Seidel et al. 2022).Hintergrund der Karte ist eine Schräglichtschummerung aus digitalen Geländedaten (Geobasis NRW)

Fig. 1
Fig. 1 Topographic map showing portions of the municipality of Bochum.Based on our field survey, springs (and small creek valleys) have been mapped (base map: Geobasis NRW).Sampling-wells and -sites are indicated.Note the small overview map with an inset of the location of the Ruhr District in Germany Abb. 1 Topographische Karte im südlichen Stadtgebiet von Bochum.Auf der Grundlage unserer Feldbegehung wurden Quellen (und kleine Bachtäler) kartiert (Basiskarte: Geobasis NRW).Probennahme-Messstellen und -Standorte sind eingezeichnet.Beachten Sie die kleine Übersichtskarte

Fig. 5
Fig. 5 Thin section images of the Schöttelchen sandstone.a Quartz grain under polarised light.b Albite with typical lamellar twinning.c A flake of hard coal embedded in the sandstone matrix.d An example of microcrystalline quartz cement.e Muscovite, a typical detrital mineral.Due to pressure solution, quartz grains are closely packed.f A calcite cement within a fracture.The black void at the top of the image is an open fracture space.g, h BSE images of an open fracture (black).Both g and h show Fe-dolomite as fracture filling mineral phase.Mg 2+ , Fe 2+, and Ca 2+ contents vary within the cement phase occluding the fracture (BSE images: Olschowsky 2020) Abb. 5 Dünnschliffbilder des Schöttelchen-Sandsteins. a Quarz unter polarisiertem Licht.b Albit mit typischer lamellarer Verzwillingung.c Ein in die Sandsteinmatrix eingebetteter Steinkohlesplitter.d Ein Beispiel für mikrokristallinen Quarzzement.e Muskovit, ein typisches detritisches Mineral.Aufgrund der Drucklösung sind die Quarzkörner dicht gepackt.f Ein Calcit-Zement in einer Kluft.Der schwarze Bereich oben auf dem Bild ist der offene Kluftraum.g, h BSE-Bilder einer offenen Kluft (schwarz).Sowohl g als auch h zeigen Fe-Dolomit als kluftfüllende Mineralphase.Die Gehalte an Mg 2+ , Fe 2+ und Ca 2+ variieren innerhalb der Zementphase, die die Kluft verschließt (BSE-Bilder: Olschowsky 2020)

Table 2
Phase transfers for ten models calculated with inverse PHREEQC modelling for shallow and deep groundwater compositions.Units in moles/kg water.Positive values indicate the phase enters the solution; negative values indicate the phase leaves the solution."-" indicate the phase is not necessary for the model Tab. 2 Phasentransfers für zehn Modelle, die mit der inversen PHREEQC-Modellierung für flache und tiefe Grundwasserzusammensetzungen berechnet wurden.Einheiten in Mol/kg Wasser.Positive Werte zeigen an, dass die Phase in die Lösung gelangt; Negative Werte zeigen an, dass die Phase die Lösung verlässt."-" gibt an, dass die Phase für das Modell nicht erforderlich ist Variants:

Table 3
Equilibrium loading of cation exchanger and maximum amounts of species transfer in the PHREEQC-model Tab. 3 Equilibrium-Beladung des Kationenaustauschers und maximale Mengen des Speziestransfers im PHREEQC-Modell aMaximum transfer in all models