Introduction

Methane is both a natural and a man-made contaminant in groundwater. Methane dissolved in drinking water is not considered to be a public health risk although medical research on this topic is limited (Jackson et al. 2011). Nevertheless, high aqueous methane concentrations affect the groundwater quality in an indirect manner. High pH together with low oxygen fugacity may cause the release of heavy metals or hazardous semi-metals from the aquifer thus posing a potential health risk.

The more evident effect of high aqueous methane concentrations is the separation of a gas phase with pressure lowering causing explosions, fires or asphyxiations. The debate on man-made methane pollution has intensified in recent years. The main reason is the increasing use of unconventional techniques in the exploitation of natural gas reservoirs. These methods are commonly based on hydraulic fracturing or fracking. The basic principle is the injection of a fracking fluid into the reservoir thus loosening the formation and releasing gas. The fluid is a mixture of water, quartz and dissolved chemicals, some of them toxic.

There is a huge amount of published information on hydraulic fracturing, the number of papers totalling 60,000 over the past 30 years (King 2012). This contrasts with the little information on hydrochemical aspects associated with the potential methane leakage during such an operation. Some papers deal with the stable isotopes of carbon, hydrogen and helium, which are meant to distinguish between man-made sources and natural occurrences (Taylor et al. 2000; Breen et al. 2007; Osborn and McIntosh 2010; Osborn et al. 2011; Stuart 2012; Jackson et al. 2013; Molofsky et al. 2013; McMahon et al. 2013). There are reports on elevated concentrations of arsenic, selenium and strontium in groundwater near fracking sites in the Barnett Shale formation of north Texas but it is not known whether the contaminants have a near-surface origin or are derived from a deep source (Fontenot et al. 2013). The elevated levels of potassium, chloride and pH near fracking sites in the Wind River formation near Pavillion, Wyoming, are rather related to mixing groundwater with fracking fluid than methane intrusion (Myers 2012).

To the best of the author’s knowledge, there are no descriptions of the chemical changes that are produced by methane gas escaping from the fracking zone and the movement of liquid associated with it. This paper is the first attempt to model the combined effects of intruding methane gas and aqueous liquid with respect to the mobilisation of hazardous trace elements.

The model site is the Damme area, northern Germany (Fig. 1). It is a real-world case except that a hypothetical leakage zone is included. This zone connects a deep Cretaceous horizon, which has been fracked in a pilot operation, and a shallow Quaternary aquifer, which carries potable groundwater. The site has typical features of other potential fracking targets in northern Germany (Andruleit et al. 2012).

Fig. 1
figure 1

TOUGHREACT model of the Damme 3 area: Vertical nodal position (z coordinates of the fresh water head; m a.s.l.) of the top layer. Location map: A Austria, B Belgium, CH Switzerland, CZ Czech Republic, DK Denmark, F France, L Luxembourg, NL Netherlands, PL Poland

Full size image

Materials and method

Simulation programs

The simulations are performed with a modified version of the TOUGHREACT 2.0 code, which calculates chemically reactive flows of multi-phase fluids (Xu et al. 2012). The programme was developed by introducing reactive chemistry into the TOUGH2 flow code (Pruess et al. 1999). The governing equations are discretised using integral finite difference in space and fully implicit first-order finite difference in time.

Methane is the non-condensible gas in the modified eco2n module of the TOUGHREACT 2.0 code. Instead of calculating the three-component flow of water, NaCl and CO2, the system water-NaCl-CH4 is used. The thermodynamic properties of the system are identical with those of the system water-NaCl-CH4 in the eco2 module of the TOUGHREACT 1.0 code (Xu et al. 2005). This early version of the code does not have an option for calculating surface complexation, which plays a vital role in contaminant-transport modelling.

Model site geology and hydrogeology

ExxonMobil carried out three fracs in the Damme 3 well, Lower Saxony, in 2008. To date, this the only fracking operation in a shale-gas deposit in Germany. The fracs are located at elevations of about −1,500, −1,400 and −1,000 m a.s.l. in the Lower Cretaceous Bückeberg formation (“Wealden facies” or “German Wealden”; Meiners et al. 2012; Sieber 2012). This limnic–brackish sequence has been deposited in the 400-km long and 100-m wide Lower Saxony basin during the Lower Cretaceous regression that effected large parts of Europe north of the Alps. The Bückeberg formation is composed of claystone, bituminous limestone, dolomitic limestone and coal seams. Its thickness is exceptional (>500 m) in the Damme 3 area. The 650-m thick Valanginian-Hauterivian claystone overlies the Bückeberg formation, followed by 220-m thick Upper Campanian through Eocene claystone and sandstone, 30- to 40-m thick Quaternary sands and, finally, drift marl of the Drenthe glacial stage (Table 1; Goens 2011; Gunzelmann and El Hamdaoui 2011). The freshwater head is located at about 39 m a.s.l. (Fig. 1).

Table 1 Set-up of the TOUGHREACT flow model
Full size table

The model comprises a 40-m thick Quaternary sequence of the Drenthe glacial stage with sand (q1) overlain by drift marl (q2) and a hypothetical leakage zone down to a depth of −969 m a.s.l. The vertical extension of the leakage zone is far beyond the height of a fracture that can be produced in a fracking operation. Andruleit et al. (2012) calculate that the injection of 4,400 m3 fracking fluid, such as the fluid volume for the uppermost frac in the Damme 3 well, can produce a 0.03-m wide fissure with a maximum extension of 350 m vertically and 540 m horizontally. Thus, the exceptional scenario in this paper implies the existence of a pre-fracking high-permeability zone above the zone of influence of the actual fracking operation. The 0.03-m wide and 4-m long leakage zone has a permeability in the range from 10−10 to 10−8 m2.

The justification for the exceptional scenario is the absence of published data that could characterise the integrity of the cap rock. The lack of relevant data, such as 3D seismic surveys and pre-fracking hydraulic drill-hole tests, contrasts with the wide scientific activity initiated by the Damme 3 pilot fracking test and following public debate (Sauter et al. 2012; Andruleit et al. 2012; Meiners et al. 2012; Gordalla et al. 2013; Hammes et al. 2013; Hübner et al. 2013; Kerschke and Schulz 2013; Lange et al. 2013; Olsson et al. 2013).

The already exceptional leaking-gas scenario is supplemented by an extremely pessimistic leaking-liquid scenario. It is assumed that fracking liquid leaks through faulty well seals at intermediate depths (300–350 m) and travels along subhorizontal pathways, e.g., along the Hauterivian-Campanian unconformity. The contaminated liquid is assumed to get into contact with rising methane gas. Of course, the likelihood that unfavourable geological conditions and technical failure coincide in the same operation is extremely small. Nevertheless, the wide use of the fracking technique with ten thousands of individual operations is a great challenge for the regulatory agencies. The best option is to consider even the most unlikely events.

Mesh generation

The model mesh calculated with the AMESH code (Haukwa 1998) has 31 layers with a total of 5,728 nodes and a vertical extension of 1,008 m (Figs. 1, 2; Table 1). The topmost part of the mesh has 8 layers and a horizontal extension of 88 × 90 m. The remaining mesh consists of a single column with 23 elements located in the centre of the mesh below layer 8.

Fig. 2
figure 2

Block diagram of the TOUGHREACT model of the Damme 3 area

Full size image

The eight topmost layers have a nodal distance of 4 m in the x direction and 4 sets of nodal distances in the y directions. The central segment is 0.03 m wide, followed by 0.1, 0.3 and 1 m wide segments in both y-directions. The remaining segments are 4 m wide. The top layer has nodal z values representing the freshwater head (Fig. 1). The second topmost layer has nodes located at 0.1 m below those of the top layer. The following six layers have a thickness of 5.8 m. The remaining mesh consists of a column with 22 layers with a thickness of 44 m and a bottom cell with a node located 0.1 m below that of the second lowermost cell. The horizontal dimensions of the column are identical to those of the overlying elements, i.e., a horizontal interface area of 0.12 m2 or 4 m in the x-direction and 0.03 m in y-direction.

The top layer, which serves to maintain constant pressure (105 Pa), temperature and salinity, exclusively consists of infinite-volume elements (1052 m3). The thermodynamic conditions of these boundary elements do not change at all. Thus, Dirichlet conditions are implemented at the boundary between the topmost and second topmost layer. The bottom cell has a volume of 103 m3. This volume is large enough for maintaining nearly constant salinity and temperature throughout the initialisation period but is simultaneously flexible enough to account for pressure adjustments. The volume of the remaining cells is calculated according to their nodal positions.

Initialisation of flow and non-reactive transport conditions

The infinite-volume elements of the top layer (1052 m3), which impose Dirichlet conditions, have a pressure of 105 Pa and a temperature of 11.5 °C. The salinity is set at 0.00005 mass fraction NaCl, which corresponds to the median Cl concentrations of potable groundwater in northern Germany (Kringel and Bahr 2006). The large-volume element (103 m3) of the bottom cell at −969 m a.s.l. imposes nearly Dirichlet conditions with respect to temperature and salinity. The temperature in this layer is 50 °C and the salinity is set at 0.16 mass fraction salt (Müller and Papendieck 1975). The intermediate layers have pre-initialisation temperature and salinity values that are interpolated according to their depths. The concentration of non-reactive tracer and the gas saturation is zero in all cells.

The initialisation is carried out in a 100-year base-case run and two sensitivity-case runs. The base-case initialisation is performed with a leakage zone having a permeability of 10−8 m2, which is close to the theoretical maximum of 10−7 m2 (Langguth and Voigt 2004). In addition, two sensitivity-case initialisations are performed with lower permeabilities (10-10 and 10−9 m2). The saturated hydraulic properties of the two remaining geological units are identical in the base case and the sensitivity cases (Table 1). Due to the lack of site-specific information, the data from a similar study of the Gorleben area (Schwartz 2012) have been assigned to the Quaternary sands (q1) and drift marl (q2).

Flow and non-reactive transport simulation

The slightly modified output of the flow and non-reactive transport initialisation run is used as initial flow condition for the TOUGHREACT base-case simulation. The CH4 gas saturation in the lowermost cell is changed from zero to 0.2. The concentration of the non-reactive aqueous tracer 1 and 2 is changed from zero to 0.001 mol/L in the layer 15 (−306 m a.s.l.) and layer 16 (−350 m a.s.l.), respectively. The volume of the lowermost cell is increased from 103 m3 to 1052 m3; thus perfect Dirichlet conditions are implemented at the boundary between the lowermost and second lowermost layer.

The capillary pressure and the relative permeability are calculated with the functions of Brooks and Corey (1964), which use the parameters λ (pore-size distribution index) and P c (bubbling pressure). The base-case λ value is 2.0. The same value is incorporated in the relative-permeability function of Corey (1954). In addition, two sensitivity cases are calculated with λ = 0.5 and λ = 6.0, thus covering the full range of experimental λ values published by Brooks and Corey (1964). The P c value, which relatively little influences the two-phase flow conditions, is that of the Apache Leach tuff reported by Rasmussen (2001) in all scenarios.

Initialisation of reactive-transport conditions

The initial solutions of the entire model are in equilibrium with the solid phases shown in Table 2 and are subject to equilibrium surface-complexation reactions with illite. The calculations are performed with an expanded THERMODDEM thermodynamic database (Blanc et al. 2012) using all aqueous species of the elements shown in Table 3. The modification consists of adding surface complexation reactions with illite and dissociation reactions of UO2.69 (solid phase) and galena with a Gibbs free energy of formation of −20.7 kcal/mol (Schwartz 2014). Furthermore, some new additions are necessary (Table 4 and 5):

Table 2 Initial mineral volume fractions and kinetic properties of the TOUGHREACT transport model
Full size table
Table 3 Initial water solution in the topmost layer of the q1 aquifer of the TOUGHREACT transport model and statistical parameters of the database AQUA
Full size table
Table 4 Equilibrium constants [log(K)] for dissociation reactions of pollutant phases added to the THERMODDEM thermodynamic database, used for TOUGHREACT simulations
Full size table
Table 5 Equilibrium constants [log(K)] for surface-complexation reactions of the TOUGHREACT transport model
Full size table

  • The solid-phase FeAs0.00001Se0.00001S1.99998 (As-Se-bearing pyrite) is added. This is a calibrated phase with respect to As and Se. The dissociation constants are identical with those of pyrite.

  • The solid-phase Cd0.00000001Cr0.000001 Ni0.000005Fe0.999994OOH (Cd-Cr-Ni-bearing goethite) is added. This is a calibrated phase with respect to Cd, Cr and Ni. The dissociation constants are identical with those of goethite.

  • Surface-complexation reactions with adsorption of Cd and Cr onto illite are added. The equilibrium constant for Cd+2 has been taken from Gu and Evans (2007). For Cr+3, the equilibrium constant for surface complexation onto hydrous ferric oxide (Scott 2010) is used.

The resulting concentrations of the trace elements in the topmost q1 layer are close to the median of the chosen reference data set. This is the hydrochemical database AQUA (Kringel and Bahr 2006), which contains the analyses of 394 samples from northern Germany, including a large range of trace elements. From these, only the trace elements are selected for this study whose median is at least 1 % of the WHO regulatory limit: As (regulatory limit 1.33 × 10−7 mol/L), Cd (2.67 × 10−8 mol/L), Cr (9.61 × 10−7 mol/L), Ni (3.41 × 10−7 mol/L), Pb (4.83 × 10−8 mol/L), Se (1.27 × 10−7 mol/L) and U (8.40 × 10−9 mol/L). As opposed to the trace components, the pH and the concentration of major elements depend on non-calibrated phases of the thermodynamic data base. Thus the intitial pH and the initial concentrations of major elements in the topmost q1 layer do not necessarily have to be close to the median of the reference data set but they do lie in the range of the 10 % quantile and 90 % quantile of the reference data set (Table 3).

The surface-complexation reactions are calculated with the double-diffusive-layer model. The site density (5.8 × 10−6 mol/m2) as well as some surface-complexation equilibrium constants are taken from a compilation of Schwartz (2014), other sources being Gu and Evans (2007) and Scott (2010). The surface area for surface complexation onto illite is a calibrated maximum value. It is assumed that the maximum amount of As, Cd, Cr, Ni, Pb, Se and U adsorbed onto illite cannot be greater than the amounts corresponding to the average concentrations in sands and sandstone. The limit is set by the average As concentration in sandstone (1.4 ppm As; Wedepohl 1969–1978), assuming a rock grain density of 2.7 g/cm3. Accordingly, the maximum absorbent surface area (base-case surface area) is 9.6 × 103 cm2/g, other configurations allowing a greater surface area. The reactive surface areas (Table 2), which are independent of the surface areas for surface complexation (adsorption), are based on Xu et al. (2012).

The initial fluid composition is determined with the calibrated mineral zones at temperatures and pressures calculated with the coupled flow module of TOUGHREACT. The pre-initialisation Cl concentration is 7.9 × 10−4 mol/L from the top of the model down to a depth of −175 m a.s.l., and then linearly increases up to 1.8 mol/L down the model at −527 m a.s.l. but remains constant down to the bottom of the model at −969 m a.s.l. The concentrations of Na are adapted in such a manner that the charge balance is maintained.

Apart from the base case, eight sensitivity cases are calculated. The initialisation of these cases is identical to that of the base case except that some sensitivity cases use different configurations with respect to surface complexation. The initialisation of these cases is performed with adsorption onto illite with base-case surface areas for surface complexation multiplied with 0.1 as well as with zero adsorption.

Reactive-transport simulation

The slightly modified output of the flow initialisation run is used as initial flow condition for the TOUGHREACT base-case simulation. The modification consists of changing the CH4 gas saturation in the lowermost cell from zero to 0.2. Another modification concerns the model mesh. The volume of the lowermost cell is increased from 103 m3 to 1052 m3; thus perfect Dirichlet conditions are implemented at the boundary between the lowermost and second lowermost layer.

Surface complexation reactions are calculated as equilibrium reactions according to the requirements of the code. In contrast, the mineral reactions are calculated as kinetically controlled reactions. The dissolution-rate parameters are taken from a compilation of Palandri and Kharaka (2004) except those for galena (Zhang et al. 2004), whereby the parameters for dolomite, smectite and uraninite are used for siderite, illite and UO2.69, respectively (Table 2). Precipitation-rate parameters are assumed to be equal to dissolution-rate parameters. The simulation period is 1, 2 and 4 days.

For the eight sensitivity cases, the simulation procedure is identical to that of the base expect that different settings with respect to surface complexation and/or kinetically controlled reactions rates are used. The sensitivity-case rate constants are the base-case rate constants multiplied by 0.1 and 0.01. The sensitivity-case adsorption scenarios refer to base-case surface areas of illite multiplied by 0.1. Another set of sensitivity cases completely neglects adsorption.

Results

Base case

The rapid upward movement of methane gas in the leakage zone is accompanied by a much less dynamic migration of the liquid phase. At the start of the simulation, methane gas is only present in the lowermost cell at −969 m a.s.l. In little less than half a day, methane gas arrives at the third topmost layer (top of the potable water aquifer) and, after 4 days, reaches the lateral boundaries of the model (Fig. 3).

Fig. 3
figure 3

TOUGHREACT 2.0 models for the Damme 3 area. CH4 gas saturation in the top layer of the Quaternary sands (q1) for the base-case permeability of the leakage zone (10−8 m2) and sensitivity case permeabilities of the leakage zone (10−9 and 10−10 m2) at the end of the 1–day, 2-days and 4-days simulation periods. The pore size distribution index is 2.0 (base case)

Full size image

The movement of liquid, which accompanies the rises of methane gas, is relatively slow. The profiles of the tracer concentrations (Fig. 4) indicate that part of the liquid column has moved upward by about 260 m at the end of the 2-day simulation period. A similar travel length can be deduced from the change of the Cl profile (Fig. 5). The travel length increases little if the simulation is continued beyond 4 days (not shown). This implies that rising methane gas that gets into contact with contaminated liquid (e.g. fracking fluid escaping through faulty well seals) is not capable of moving the aqueous contaminants more than 300 m vertically.

Fig. 4
figure 4

TOUGHREACT 2.0 models for the Damme 3 area using the base-case permeability for the leakage zone (10−8 m2) and sensitivity-case permeabilities for the leakage zone (10−9 and 10−10 m2) in combination with the base-case pore-size distribution index of 2.0. Concentration of the non-reactive tracer 1 and 2 in groundwater in the leakage zone and overlying layers at the beginning and end of the 1-day, 2-day and 4-day simulation periods

Full size image
Fig. 5
figure 5

TOUGHREACT 2.0 base-case model for the Damme 3 area. Temperature, CH4 gas saturation and chemical composition of the groundwater in the vertical leakage zone and overlying layers at the beginning and end of the 2-days simulation period

Full size image

The intruding methane causes the acidity and oxygen fugacity to decrease (Figs. 6 and 7). Goethite (incorporating Cd, Cr and Ni) dissolves with decreasing acidity and oxygen fugacity whereas carbonates and sulphides (incorporating As, Cd, Pb and Se) and UO2.69 precipitate.

Fig. 6
figure 6

TOUGHREACT 2.0 base-case model for the Damme 3 area. CH4 gas saturation, pH, concentrations of O2,aq and Cr in groundwater of the top layer of the Quaternary sands (q1) at the end of the 1-day, 2-day and 4-day simulation periods

Full size image
Fig. 7
figure 7

TOUGHREACT 2.0 base-case model for the Damme 3 area. Vertical W–E sections with CH4 gas saturation, pH and concentrations of O2,aq and Cr in groundwater at the end of the 2-days simulation period. Horizontal exageration = 200 %

Full size image

With increasing pH, the positively charged species Cd+2, Ni+2, Pb+2 and UO +22 as well as the neutral aqueous species As(OH)3 are adsorbed by surface complexation onto illite in the relevant pH range (7–9). The elements whose dominant aqueous species are negatively charged in this pH range (Cr and Se) are released with increasing pH. The desorption effect partly is counterbalanced by the precipitation of Se-bearing pyrite but enhanced by the dissolution of Cr-bearing goethite. Thus, the intrusion of methane is accompanied by a relatively strong decrease of the concentration of As, Cd, Ni, Pb and U as well as a relatively strong increase of Cr.

In the upper layer of the q1 aquifer, the distribution of the hazardous elements is mainly controlled by the distribution of methane gas, whereas in the lower layer, there is the additional effect of the intruding Cl-enriched aqueous phase (Fig. 5). This is only evident in the case of elements that form strong aqueous complexes with Cl (Cd, Ni and Pb). Thus, the depletion effect caused by the acidity and oxygen-fugacity decrease is only partly counterbalanced by the enhancement effect caused by the Cl component of the groundwater that accompanies the rise of methane gas. In a scenario where methane and Cl are the only additional components, the concentrations of hazardous trace components stay well below the regulatory limit except that Cr is quite close to it (9.61 × 10−7 mol/L or 50 µg/L).

Sensitivity cases

The influence of changing flow conditions is shown by reducing the base-case permeability of the leakage zone (10−8 m2) by one and two orders of magnitude (Fig. 3). The low-permeability case (10−10 m2) implies that methane gas needs about 2 days to arrive at the top of the q1 aquifer as opposed to half a day in the high-permeability case (10−8 m2). The size of the plume in the low-permeability case at the end of the 4-day simulation period is smaller than that of the high-permeability case after a half-day simulation period (not shown).

Figure 8 shows the flux calculations performed with the two different versions of the code, i.e. the original eco2 module of TOUGHREACT 1.0 and the modified eco2n module of TOUGHREACT 2.0 (see “Simulation programs”). The flux of methane gas at the top of the leakage zone is almost identical in both types of simulations. The flux of liquid shows some differences, which tend to decrease with decreasing permeability (not shown). Note that the differences between the two types of simulation are minor compared to the differences shown by a similar code intercomparison: the CO2-water system modelled by the STOMP code and different versions of the TOUGH2 code with the eco2 module (Pruess et al. 2002; Fig. 4.6).

Fig. 8
figure 8

Comparison of the TOUGHREACT 1.0 code and TOUGHREACT 2.0 code (modified). Simulation of the flux of liquid and methane gas at the top of the leakage zone with a permeability of 10−8 m2 and pore-size distribution index of 2.0 (both are base-case values)

Full size image

The permeability variations strongly influence the horizontal distribution of the reactive aqueous species, which correlates with the CH4 gas saturation (not shown because a similar relationship is exhibited on the time scale; see Fig. 6). The influence of permeability variations on the vertical distribution of aqueous species is very minor (Fig. 4). The same goes for variations of the pore-size distribution index (Fig. 9).

Fig. 9
figure 9

TOUGHREACT 2.0 models for the Damme 3 area using the base-case pore-size distribution index (λ = 2.0) as well as sensitivity-case values of 6.0 and 0.5. The permeability of the vertical leakage zone is 10−8 m2 (base case). The concentration of the non-reactive tracer 1 and 2 in groundwater in the leakage zone and overlying layers at the beginning and end of the 2-day simulation period is shown

Full size image

The influence of changing transport conditions is shown by eight sensitivity cases, which are calculated for a simulation period of 2 days. These cases have in common that the chemical changes caused by the methane intrusion are less dynamic than those of the base case, which implies maximum values for rate constants and adsorption/desorption.

Reducing the degree of adsorption/desorption by reducing the surface area of illite or neglecting adsorption altogether reduces the degree of depletion of Ni and reduces the degree of enrichment of Cr and Se (Fig. 10). Similarly, the degree of depletion of As, Cd, Pb and U is reduced with decreasing adsorption (not shown). Note that the effect of competition for surface complexation sites is only calculated for those eight aqueous species that are defined by the reactions in Table 5. Increasing the number of competing species is theoretically possible but is hampered by the lack of suitable experimental data, in addition to convergence problems associated with such a procedure. The practical consequences are nil because increasing the degree of competition has the same effect as the reduction of the surface area of illite in the sensitivity case calculations.

Fig. 10
figure 10

TOUGHREACT 2.0 transport models for the Damme 3 area using the base-case flow model. Concentration of Cr, Ni and Se of groundwater in the vertical leakage zone and overlying layers at the end of the 2-day simulation period. a–c Cases with adsorption and base-case surface areas whereby the red symbols refer to the base case (base-case rate constant for kinetically controlled reactions) and the green and black symbols refer to sensitivity cases (base-case rate constants multiplied by 0.1 and 0.01). d–f Sensitivity cases with adsorption, and base-case surface areas multiplied by 0.1. g–i Sensitivity cases without adsorption

Full size image

Reducing the rate constants reduces the degree of depletion of Cd, Pb and Ni and reduces the degree of enrichment of Cr provided that surface complexation is considered (only shown for Cr and Ni in Fig. 10). The effects on As, Se and U concentrations are minor because the dissolution of As-Se-bearing pyrite and UO2.69 only plays a minor role.

Discussion and conclusion

The geological risks involved in a fracking operation in methane-bearing shale are minor in comparison to the technical risks. Nevertheless, the geological hazards deserve attention. This is especially the case when methane gas risks to rise from the fracking site towards the surface. The pathway may be a fracking-induced opening, pre-existing fault zone or combination of both. Chemical reactions take place between methane, Cl-enriched groundwater and aquifer.

The methane intrusion causes the pH to increase from about 7 to 9. This is accompanied by surface complexation reactions that release elements whose dominant aqueous species are negatively charged in this pH range (Se and Cr). The desorption effect may be counterbalanced by precipitating sulphides (e.g. Se-bearing pyrite) and there may be an enhancement effect by dissolving oxides (e.g. Cr-bearing goethite). The effect of increased salinity on the distribution of elements that form strong chloride complexes (Cd, Ni and Pb) is completely offset by adsorption reactions. These are theoretical considerations. Ideally, pre-fracking hydrochemical baseline surveys and post-fracking monitoring should be combined. Short of specific studies, it is suggested that the TOUGHREACT model of this paper is used as a substitute for empirical studies. At best, this paper may stimulate combined empirical and theoretical research on sites with a hazard of methane pollution.

The unlikely case of technical failure in combination with unfavourable geological conditions would have serious consequences. The constituents of the fracking fluid impose a special risk if leakage through faulty well seals takes place less than 300 m below the potable-water aquifer. It cannot be excluded that, under unfortunate circumstances, rising methane gas gets into contact with groundwater contaminated by fracking liquid escaping through faulty well seals. Rising methane gas can induce the contaminated liquid to travel up to 300 m vertically. The sealing of the fracking wells in this particularly critical depth range, down to 300 m below the potable-water aquifer, should be optimal. Even special risk-reducing techniques such as double or triple well sealing deserve to be considered.