Hydraulic conductivity of fractured upper crust: insights from hydraulic tests in boreholes and fluid-rock interaction in crystalline basement rocks

Transcription

1 Geofluids (2015) 15, doi: /gfl Hydraulic conductivity of fractured upper crust: insights from hydraulic tests in boreholes and fluid-rock interaction in crystalline basement rocks I. STOBER 1 AND K. BUCHER 2 1 Karlsruhe Institute of Technology KIT, Institute of Applied Geosciences, Karlsruhe, Germany; 2 Mineralogy and Petrology, University of Freiburg, Freiburg, Germany ABSTRACT The permeability (j[m 2 ]) of fractured crystalline basement of the upper continental crust is an intrinsic property of a complex system of rocks and fractures that characterizes the flow properties of a representative volume of that system. Permeability decreases with depth. Permeability can be derived from hydraulic well test data in deep boreholes. Only a handful of such deep wells exist on a worldwide basis. Consequently, few data from hydraulically tested wells in crystalline basement are available to the depth of 4 5 km. The permeability of upper crust varies over a very large range depending on the predominant rock type at the studied site and the geological history of the drilled crystalline basement. Hydraulic tests in deep boreholes in the continental crystalline basement revealed permeability (j) values ranging over nine log-units from to m 2. This large variance also decreases with depth, and at 4 km depth, a characteristic value for the permeability j is m 2. The permeability varies with time due to deformation-related changes of fracture aperture and fracture geometry and as a result of chemical reaction of flowing fluids with the solids exposed along the fractures. Dissolution and precipitation of minerals contribute to the variation of the permeability with time. The time dependence of j is difficult to measure directly, and it has not been observed in hydraulic well tests. At depths below the deepest wells down to the brittle ductile transition zone, evidence of permeability variation with time can be found in surface exposures of rocks originally from this depth. Exposed hydrothermal reaction veins are very common in continental crustal rocks and witness fossil permeability and its variation with time. The transient evolution of permeability can be predicted from models using fictive and simple starting conditions. However, a geologically meaningful quantitative description of permeability variation with time in the deeper parts of the brittle continental crust resulting from combined fracturing and chemical reaction appears very difficult. Key words: brittle ductile transition, deep well, hydrothermal, permeability, veins, well test Received 13 January 2014; accepted 20 August 2014 Corresponding author: Ingrid Stober, Karlsruhe Institute of Technology KIT, Institute of Applied Geosciences, Adenauerring 20b, Karlsruhe D-76131, Germany. ingrid.stober@kit.edu. Tel: Geofluids (2015) 15, INTRODUCTION Features of brittle deformation, such as fractures, faults, joints, and veins, are the principal water (fluid)-conducting structures in crystalline basement rocks and provide the dominant conduits for fluid flow in the brittle upper continental crust. Fluid flow in this environment can be described by the flow law of Darcy. Flow is driven by a hydraulic gradient and controlled by the prime parameters of advective fluid, solute, and heat transport in fractured crystalline rocks, namely permeability and porosity. Depending on the scale of interest, different methods are used to determine these two parameters. These include laboratory measurements (e.g., Berckhemer et al. 1997), fracture analysis (Caine et al. 1996; CFCFF 1996; Jakob et al. 2003; Mazurek et al. 2003; Konzuk & Kueper 2004), geophysical modeling of heat flow data (e.g., Hayba & Ingebritsen 1994), and in situ testing of boreholes (e.g., Cooper & Jacob 1946; Ferris 1951; Butler 1998; Nielsen 2007; Peters 2012). Well tests provided permeability data from the crystalline basement to depths of 5 km and offer insights into the permeability 2014 John Wiley & Sons Ltd

2 162 I. STOBER & K. BUCHER structure of the crust and its variation with depth and lithology. It is important, however, to critically analyze and evaluate permeability data derived from well tests reported in the literature before they can be used (or discarded) in large-scale models of the permeability structure of the crust. In this study, we address several aspects and difficulties in deriving permeability data from hydraulic well tests. Porosity and permeability are subject to unceasing changes in the brittle upper continental crust. Reactive fluid flow along the fractures continuously modifies the aperture and surface structure of the fractures mostly by chemical dissolution and precipitation reactions (Weisenberger & Bucher 2011; Bucher et al. 2012; Alt-Epping et al. 2013). After initial porosity and permeability generation by creation of a fracture, the permeability tends to increase during a certain period. Later, permeability and porosity decreases due to precipitation of secondary minerals along the fractures until they become completely sealed. This fracture reaction seal process of reaction vein formation may occur in several cycles and cause incremental vein growth. Although there are ample field observations that imply the significance of the described permeability time relationship in fractured upper crust (e.g., Bucher-Nurminen 1981), the time-dependent permeability evolution cannot normally be recorded by hydraulic well tests because of incompatible timescales of the fracture reaction seal process (Bucher-Nurminen 1989) and the well test methods. Nevertheless, we try to illustrate in this study structures characteristic of transient permeability in fractured upper crustal rocks. Fluid flow in fractured upper crustal aquifers is generally driven by hydraulic gradients, which may result from a number of different feasible causes and imbalances including topography, thermal, and chemical disequilibrium. Pumping or injection tests carried out in boreholes are artificially induced hydraulic gradients as forcing for fluid flow. Hydraulic tests provide data on the hydraulic properties and the nature of aquifers and the permeability structure of the upper crust including the depth variation of permeability. Hydraulic well tests also have the potential to identify and locate faults or fault systems in the crust. The tests can quantify the hydraulic conductivity of the faults if it is different from that of the fractured rock matrix. Fault conductivity can be lower than, equal to, or higher than that of the rock matrix. Examples of low-permeability faults have been presented in Stober & Bucher (2005a) and Stober et al. (1999). Important and frequently overlooked driving forces for fluid flow are tidal forces that prevent fluids in fractured systems from ever becoming stagnant and chemically inactive. We also illuminate the effect of tidal forces on flow and reaction of fluids in fractured basement rocks. PERMEABILITY, SIGNIFICANCE IN FRACTURED BASEMENT ROCKS The permeability j [m 2 ] of a volume of fractured crystalline rocks forming the upper continental crust relates to the structure and connectivity of the fluid-filled pore space. The permeability controls, together with the viscosity and density of the fluid, the ability of the volume of rock to conduct a fluid phase. Permeability is, together with the fluid properties, a decisive parameter controlling fluid flow in brittle rocks. It can be retrieved from the transmissivity T [m 2 sec 1 ] calculated from hydraulic test data. From the experimentally derived transmissivity T, the hydraulic conductivity K [m sec 1 ] can be computed. Quantifying T and K from pressure time data obtained in well tests requires geological expertise and ideas about the geological and hydraulic structure of the subsurface. The parameter K is not measured data; it is ultimately the result of a modeling process. The Darcy flow law describing fluid flow through porous media can be written as q~ ¼ K ½ Š 1 gq rp ð1aþ where q~ is the specific discharge vector per cross-sectional area [m 3 sec 1 m 2 ] with the components q x, q y, q z,[k] [m sec 1 ] the tensor of hydraulic conductivity, q [kg m 3 ] the density of the fluid, g [m sec 2 ] the acceleration due to gravity, and P [kg m 2 sec 2 or Pa m 1 ] the vector of the pressure gradient. The one-dimensional form of Darcy flow is q ¼ K 1 DP ð1bþ gq D1 with ΔP [Pa] the pressure difference along the baseline Δl [m]. In hydraulic tests, Q [m 3 sec 1 ] corresponds to the rate of pumping fluid from or injecting fluid into a wellbore. It is a parameter that can be varied during the test. Q is related to q in Eq. 1b by Q/A = q with A [m 2 ] the cross section perpendicular to flow. The imposed input signal Q results in a pressure response ΔP by the conductive system that can be measured with instruments. Data analysis of hydraulic test data requires a hydraulic aquifer model. For instance, during the radial flow period in a homogeneous isotropic aquifer, the hydraulic conductivity K is proportional to the pumping or injection rate and the inverse of the drawdown and the thickness of the tested formation in a simple and ideal case. This means, that the derived hydraulic conductivity K depends on the chosen concept for aquifer characterization. The hydraulic conductivity K, and hence also the permeability j, is not a parameter that can be measured directly and unequivocally (like length with a measuring tape or a folding ruler).

3 Hydraulic conductivity and fluid flow 163 The permeability j [m 2 ] can then be derived from hydraulic conductivity K: j ¼ K l ð2þ gq with the viscosity of the fluid l [kg m 1 sec 1 or Pa sec] and K, g, j and q from above. To better understand the significance and meaning of both the hydraulic conductivity K and the derived permeability j, it can be helpful to look behind the routine of hydraulic well testing. How are such hydraulic tests performed, how is the wanted parameter derived from the measured pressure data, and what kind of compulsory assumptions need to be made during data processing? This analysis may help evaluating the strength and significance of the parameter hydraulic conductivity, but also disclose its weaknesses. Because the subject of interest in this special volume of geofluids is the permeability structure of brittle continental crust, hydraulic tests in deep boreholes are of prime significance. At deep drilling sites, additional monitoring wells are typically not available. The deep well is simultaneously used as testing and as monitoring well. There is a generous repertoire of hydraulic tests tailored and fitted to the specific application, environment, and problem definition (and budget; Stober 1986; Krusemann & de Ridder 1991; Horne 1996). Some continuously record the pressure directly in the tested section (at depth); in other tests, pressure is measured relatively close to the surface. In many tests, the water level in the wellbore is measured as a proxy for pressure rather than the pressure itself. If no pressure data are available from the depth of interest, convoluted corrections must be applied to nearsurface pressure or water level data. The flow rate Q can be varied in hydraulic tests in many ways. Fluid can be pumped or injected at a constant rate during the entire test period; however, pumping rates can also be varied stepwise according to a detailed protocol. The flow rate may also be continuously adjusted to maintain a constant drawdown or recovery of the water level in the wellbore. In the following discussion, we presuppose that all tests are performed according to the state of the art, that all data correction is precise and correct, and that data evaluation and analysis is based on a hydraulic model that correctly reflects the geological situation underground. Test duration varies from a few minutes to several weeks. The test produces information on the hydraulic properties of the subsurface to a certain distance from the wellbore. This distance may be named radius of influence. The radius depends strongly on the test duration, the type of test method, and the hydraulic properties of the rocks. Moreover, the radius of influence also depends on the technical details of the special testing tools and the testing procedure. The hydraulic properties of the rocks in the vicinity of the wellbore are commonly altered by the drilling process itself or by drilling mud and added chemicals. Short-term hydraulic tests will not be able to see beyond this alteration zone near the wellbore. An example of the consequences of hydraulic tests with small radii of influence is the 4444-m-deep wellbore Urach 3 (SW Germany). Urach 3 was drilled, as a research borehole for deep geothermal energy utilization. From 1604 to 4444 m below surface, the bore is in granite and gneiss of the Variscan basement. The research borehole provided an unusually long series of water table measurements over a period of 13 years. These Urach 3 data have been used to decipher the hydraulic properties of the basement, the chemical composition of the deep fluid, and the interaction of this fluid with the rock matrix (Stober & Bucher 2004). Urach 3 has also been used for a large number of diverse hydraulic tests studying the influence of test methods and test duration on derived hydraulic properties. For example, long-term injection tests have been repeatedly performed, along with slug tests of short duration. A comparison of data from the two different types of tests (slug test and log-term injection test) shows that the hydraulic conductivity around the wellbore is higher than of the pristine formation by a factor of 7 (Stober 2011). Flow rate and pressure variations during injection tests can be delimited to the capacity of the formation. Derived parameters will then reflect the intrinsic properties of the formation. If tests are run outside these conditions, fracture apertures may artificially increase, and elastic reactions of the formation alter the conductivity of the system during the test. The derived parameters reflect the artificial properties created by the test itself. Injection tests in the Urach 3 research well, for example, induced elastic reactions by the gneissic basement at well head pressures above 170 bar, resulting in a significantly increased conductivity (Stober 2011). In tectonic active areas, injection tests may reduce effective stress and frictional strength of fractures. The induced stress reduction by shearing can result in an irreversible and permanent increase in the hydraulic conductivity (Evans 2005). Consequently, injection tests performed at an overpressure that is too high for the tested bedrock alter the natural hydraulic properties of the tested rock volume. Hydraulic parameters derived from such tests do not reflect the properties of the natural system. Hydraulic tests may test the entire open hole or segments of it, or the section with perforated casing. Often, packers or other technical tools are used to isolate certain sections of the wellbore for testing. The tested sections typically vary from several hundred to only a few meters. The prime result of processing and analyzing the data of a hydraulic well test is the so-called transmissivity of the tested section and not directly the hydraulic conductivity or the permeability. The transmissivity T [m 2 sec 1 ]isan

4 164 I. STOBER & K. BUCHER integral quantity characterizing the hydraulic properties of the entire tested section. If the tested section H [m] is homogeneous and isotropic, the hydraulic conductivity can be computed from the determined transmissivity: K ¼ T =H ð3þ The transmissivity of tested heterogeneous and anisotropic stratified formations can be written as an integral over all contributions: Z H T ¼ Kdh 0 ð4þ The hydraulic conductivity K [m sec 1 ] can be calculated from the experimentally derived transmissivity T from Eqs 3 and 4. Fluid flow in brittle crystalline basement rocks occurs along fractures and faults, and along lithological contacts and other discrete 2D structures. The water-conducting features form an interconnected network with characteristic hydraulic properties (Mazurek 2000; Mazurek et al. 2003; Stober & Bucher 2007a). Water-conducting structures can be regularly or unevenly distributed. Consequently, at some drilling sites, the data can be interpreted as hydraulic response of a homogeneous isotropic aquifer. At other drilling sites, other concepts may appear more appropriate. The choice of the concept depends on the amount of available geological information. If not much is known about the geological structure, the derived parameter can be correct if the structure is simple or incorrect if the structure is more complex than has been recognized. Such considerations are not made during planning of hydraulic tests, and the hydraulic conductivity is computed from Eq. 3. This practice may lead to serious misjudgments of the hydraulic conductivity and permeability of fractured rocks. The following considerations may illustrate the problem: A hydraulic test performed in a 600-m-thick open hole of a deep wellbore resulted in a measured transmissivity T = m 2 sec 1. The calculated hydraulic conductivity (Eq. 3) T/H = K = m sec 1 appears feasible for fractured basement rocks. However, the main water inflow to the open hole can be restricted to a 10-m-thick very strongly fractured fault zone. The resulting hydraulic conductivity for the fault zone is K = m sec 1, 60 times higher than the first value. Strongly channeled fluid flow is not uncommon in fractured basement. Such zones can be identified and localized by geophysical well logging. The example shows that errors and misinterpretations can increase with increasing length of the tested section. On the other hand, the example also shows that the selection of an invariant test length can be very difficult and the result questionable. The high K-value may characterize the conductive fracture zone. The lower K-value, however, may not characterize the entire system of fractured rocks and highly conductive fault zones. Geophysical well logging methods such as flow meter logs, conductivity logs, and temperature logs help to better characterize heterogeneous hydraulic properties of tested formations. Prominent inflow zones can be identified and separately tested using packer systems. The methods may produce a quantitative vertical permeability profile. However, such elaborate and costly test designs are rarely used (typically in the context of development of repositories for nuclear waste, for example, Nagra 1985, 1992). Normally, one relies on T/H data to derive hydraulic conductivity K and permeability j. It is thus evident that conductivity K and permeability j in the end do not represent true and absolute values. The derived permeability j of the continental crust, for instance, at a given point of interest, strongly depends on the details of the test design of the hydraulic test used for deriving the parameter. It may be useful and representative, but the contrary may be true as well if the basement is hydraulically complex and heterogeneous and the design of the test is unable to adequately resolve the heterogeneity, or the tests are inappropriate to derive a meaningful permeability at a point in the bedrock. However, we do not propose to give up measuring properties by experiments, but rather to highlight possible pitfalls of the methods and of uncritically using published permeability data for modeling projects. The hydraulic conductivity K characterizing the hydraulic properties of the bedrock has two components. Firstly, it describes the properties of the fracture network expressed by the permeability of the rock volume, and secondly, it depends on the properties of the fluid itself. This follows from Eq. 2, which may be rewritten to highlight this aspect of K and j. K ¼ j gq ð5þ l It can be seen from Eq. 5 that hydraulic conductivity data derived from well tests depend on the properties of the fluid residing in the fracture porosity. If one is interested in the permeability of the tested rocks at a point of interest, the density q and the viscosity l of the fluid at that point must be known (see Eqs 1a,b and 2 and definition of the parameters there). The fluid parameters q and l depend on the temperature, the pressure, the mineralization, and the gas content of the fluid. This means that hydrochemical analyses must be available from sampling points at depth, in addition to depth and temperature data. The factor l/gq in Eq. 2 converts hydraulic conductivity to permeability. It has the value m sec at 20 C for pure water, m sec at 180 C for pure water, and m sec at 180 C for 100 g kg 1 NaCl solution (using l and q data from H afner et al. 1985; Sun et al. 2008; Wagner &

5 Hydraulic conductivity and fluid flow 165 Kretschmar 2008). Particularly, the viscosity of the fluid in the example above is markedly different in hot and highly mineralized fluid from that of pure water. Pumping tests in deep thermal wells can produce hydraulic conductivity values up to 8 times higher than would be obtained under cool low-salinity conditions, independent of the permeability structure of the formation itself, just as a result of fluid properties. Derived permeability may be erroneously high if fluid properties are unknown or not adequately considered. The opposite effect can be recognized in injection tests, which normally use cool low mineralized surface water as injection fluid. The correct conversion of K to j data requires fluid data from samples representative of the tested section. Collecting and analyzing fluid samples from deep wells, particularly if the fluid contains high amounts of dissolved gases, is technically challenging and expensive, and therefore, fluid data are often not available. The permeability of rocks can also be measured on drill cores in the laboratory. It is difficult or impossible to correctly represent fractures, faults, and larger cavities in core samples. The laboratory-measured permeability typically characterizes the property of the unfractured rock matrix. In general, it is significantly lower than the permeability of large volumes of fractured and faulted basement derived from well tests. The above discussion of the parameters hydraulic conductivity and permeability relates to well tests in deep boreholes that are the prime source of K and j data. The technical details of such tests and their cross-links to the geology at the drilling site have specific consequences for the derived data. The user of these data must be aware of these aspects, evaluate data carefully, and not adopt permeability data uncritically. Long-term pumping tests in deep boreholes may be well suited to detect the hydraulic properties of fractured basement. Under favorable circumstances, long-term tests can identify, locate, and characterize large-scale hydraulic structures (so-called hydraulic boundaries). An excellent example is the highly developed pumping test in the 4-km-deep pilot hole of the continental deep drilling project (KTB) in Germany (Stober & Bucher 2005a). The test produced a reliable value for the hydraulic conductivity of homogeneous fractured continental basement at 4 km depth K = m sec 1, which results in a characteristic permeability j = m 2 given the fluid properties at depth (120 C, TDS = 62 g kg 1, l = Pa sec, q = kg m 3 ; mol weight NaCl = g mol 1 ). The fluid temperature and the salinity have been taken from Stober & Bucher (2005a), the dynamic viscosity of the fluid at 120 C and 1 molal salinity comes from Kestin et al. (1981), and the density of a 1 molal salt solution at 120 C comes from Rogers & Pitzer (1982). This value of the permeability j = m 2 may be considered a representative value for fractured continental crust at about one-third of the distance from the surface to the brittle ductile transition zone in moderately warm crust (MOHO temperature C). The value is in excellent agreement with permeability data from continental crust published by Ingebritsen & Manning (2010). Their permeability data have been compiled from continental crust worldwide and have been derived by various geophysical tools and methods but not from hydraulic well tests. Our j = m 2 value is closer to the new j versus depth fit of Ingebritsen & Manning (2010) than to their previous fit (Ingebritsen & Manning 1999). Hydraulic testing of the KTB wellbore discovered a hydraulic boundary with a lower conductivity than the fractured basement. The seal appears to be located at a distance of several hundred meters from the wellbore at 4 km depth. The hydraulic boundary was identified as a fault and correlated with a major fault zone known from surface geology. Similar water-tight hydraulically prominent faults have been identified at other localities in the central European crystalline basement (Stober et al. 1999). Both examples show that prominent fault systems in the basement are not necessarily characterized by a higher conductivity than the surrounding fractured basement. PERMEABILITY AND FLUID FLOW IN THE CRUST Hydraulic tests in wellbores to 5000 m depth worldwide reveal a remarkable variation of hydraulic conductivity of the crystalline basement from to 10 4 m sec 1. The upper 1000 m are generally characterized by the higher values but also by a greater variance (Stober 1996; Stober & Bucher 2007a). The mean variation of the hydraulic conductivity decreases rapidly with increasing depth. The decrease in the permeability j with depth has been derived from well tests in the basement of SW Germany, NE France, and N Switzerland (Stober & Bucher 2007a,b) and can be described by log j ¼ 1:38 log z 15:4 ð6þ with z the depth in km and j in m 2. The described decrease in the permeability with depth (Eq. 6) is mainly controlled by the properties of biotite gneiss and other metamorphic rocks because test data for greater depths were mostly available from wellbores in gneiss. We suspect that the decrease in the hydraulic conductivity in granitic basement would be slower with depth, because of contrasting mechanical properties of mica and feldspar (see above and in Stober & Bucher 2007a). At our studied sites, granitic basement has a higher permeability at a given depth and tectonic environment than gneissic basement (given that granite is dominated by feldspar and quartz and gneiss typically contains significantly more mica than granite).

6 166 I. STOBER & K. BUCHER Ingebritsen & Manning (1999) in their geophysical study of terrestrial heat flow derived a surprisingly similar powerlaw function for the permeability decrease with depth. Their function predicts at 4 km depth in crystalline basement that log j = 15.9 m 2. Ingebritsen & Manning (2010) compared two k z relations: (i) for tectonically active crust, that is, their 1999 curve; and (ii) another curve for disturbed crust that included hypocenter migration, EGS, and other sources of data. The new function predicts a permeability for disturbed crust at 4 km depth (KTB pilot hole) of log j = 13.5 m 2 (Eq. 12 Ingebritsen & Manning 2010) in contrast to our derived value of log j = 14.9 m 2 at the KTB site. The decrease in the hydraulic conductivity and permeability with depth has, to our knowledge, never been reported and demonstrated for a single discrete deep borehole (>1000 m) in the crystalline basement. In deep wellbores, adequate hydraulic tests at different depths over sufficiently long test intervals consistent with the requirements of a correctly chosen REV are not normally available. Urach 3 is an exceptionally deep research bore in fractured crystalline basement of SW Germany. Several hydraulic tests at different depths and a total test interval of more than 2000 m produced unique hydraulic data. The basement belongs to the so-called Moldanubic zone of the central European continental crust. The basement consists of mafic biotite hornblende gneiss (amphibolite), migmatitic gneiss, quartz diorite, biotite cordierite gneiss, and other high-grade metamorphic rocks. The unaltered high-grade rocks contain zones of hydrothermal alteration. The permeability of the basement is related to a fracture network of variable orientation. There are open fractures functioning as water-conducting structures and sealed impervious fractures. The fracture density is about 2 3 open fractures per meter (Dietrich 1982). In the years , the wellbore was drilled to 3334 m (bottom hole at that time) with a 14-m-thick open hole section. The casing has been perforated at different sections, and the accessible rock has been hydraulically tested. The wellbore has been deepened from time to time (for different reasons) in the following years. At each increment, the newly accessible rocks have been hydraulically tested. In 1992, the wellbore reached its present depth of 4444 m (Stober 2011). The incremental drilling history of the Urach 3 bore offers a unique opportunity to evaluate the variation of the hydraulic conductivity with depth (Fig. 1). Transmissivity data have been derived from pumping and injection tests at 1.77, 3.30, 3.35, 3.88, and 3.97 km depth, respectively. The hydraulic conductivity (T/H) of fractured basement gneiss decreases systematically with depth z. At 3.97 km, the hydraulic conductivity is about 2000 times lower than at 1.77 km. The depth z log (T/H) data Depth [km] Log (T/H) [m sec 1 ] Fig. 1. Log (T/H) data from well tests at five different depths in the Urach 3 borehole in gneiss of the Black Forest crystalline basement. follow a linear trend (Fig. 1). The log (T/H) versus log z data can be regressed to a linear equation: logðt =H Þ¼ 8:748 z 4:465 ðr 2 ¼ 0:976Þ ð7þ The log-linear Eq. 7 can be converted to Eq. 8 expressing the depth dependence of the permeability in crystalline basement rocks (here using data for a fixed salinity of 1 molal NaCl, 120 C and 300 bar). log j ¼ 8:748 z 12:0 ðr 2 ¼ 0:976Þ ð8þ log j ranges from at 1.77 km depth to at 4 km depth. Note that the permeability of gneiss basement at the Urach 3 site is about 400 times lower than that of the basement at the KTB site at the same depth of 4 km. In general, the hydraulic conductivity of granite is higher than that of gneissic basement in areas of strong and young deformation (e.g., the three-country corner of Germany Switzerland France). In tectonically inactive areas, fracture density and, consequently, the hydraulic conductivity of large volumes of granite can be very low (Stober & Bucher 2007a). With increasing depth, the vertical stress component usually exceeds the horizontal stress components (Brown & Hoek 1978) with the result that the open water conducting fracture system changes from predominantly horizontal to mostly vertical orientation. Vertical fluid flow and fluid exchange becomes dominant at depth, particularly in areas with significant topography (Fig. 2). The steep fluidconducting fractures support deep fluid circulation systems

7 Hydraulic conductivity and fluid flow 167 Fig. 2. Schematic water flow systems in a crystalline basement area with moderate topography (e.g., Black Forest basement). The permeability contrast between granite and gneiss has a strong control on the flow paths, in addition to topography and fault structures. The figure also illustrates changing water composition along flow. (Stober et al. 1999; Stober & Bucher 2004; Bucher et al. 2012). Hydrothermal alteration zones along vertical fractures document the existence of significant vertical permeability (Lee et al. 2011; see also section on reactive fluid flow below). One notable exception has been described from the Blancket well (Australia), where the vertical pressure component becomes lower than the horizontal component at depths below 1800 m. Consequently, fractures and faults have a preferred horizontal orientation in the 4000-m-deep stimulated geothermal Habanero 4 well (Bendall et al. 2014). Hydrochemical and isotope data suggest that thermal spring waters represent upwelling deep waters (Chebotarev 1955; Toth 1962; Stober 1995; Allen et al. 2006; Bucher et al. 2009). The conclusions from chemical data are supported by the evidence from numerical modeling (Forster & Smith 1989; Kukkonen 1995). Topography-driven hot water flow from several thousand-meter-deep sources has been reported (Toth 1978; Bucher et al. 2009). The circulation systems depend on the existence of steep fault structures that must merge into high-conductivity fault zones allowing for rapid upward flow of deep water (e.g., Baden-Baden; Stober 1995; Sanner 2000; Fig. 2). The final temperature of the deep fluid reaching the surface depends, among other factors, on the flow velocity, which is proportional to the hydraulic conductivity for a given hydraulic gradient. Typically, major fault zones with vertical displacements of hundreds or thousands of meters tend to have an inner, low-permeability central zone mantled by high-permeability rims on both sides (Mazurek et al. 2003). Topography-driven deep water circulation in the Black Forest region (Germany) results in the outflow of warm saline deep water into the freshwater aquifer of the Quaternary fill of the drainage valley (Kinzig valley). The natural contamination of the freshwater by deep salty fluids can be traced as a NaCl plume (Ohlsbach Plume) for a long distance into the Quaternary gravel aquifers of the river Rhine plain (Stober et al. 1999). The deep saline warm water flows upward along a steep fault system forming the boundary of the Black Forest topographic high and the rift system of the river Rhine valley (Fig. 3). Although the rift escarpment represents a generally extensional environment, strike-slip movements along the faults form fine-grained fault gauge, which clogs permeability, thus forming hydraulic barriers. However, many of the Rhine rift boundary faults are highly transmissive and could channel fluid flow (e.g., Bruchsal and Landau; Aquilina et al. 2000; B achler 2003). In the case of the Ohlsbach plume (Fig. 3), deep water is forced to flow upward by the hydraulic boundary fault through fractured granitic rocks that have similar conductive properties as the area of recharge. The flow velocity is slow so that the water reaches the surface with only 27 C although the reservoir at >3000 m depth has a temperature of close to 100 C (Stober et al. 1999). If upwelling deep water from similar reservoirs follows a highly permeable fault zone such as in Baden-Baden, the outflow temperature is as high as 68 C although the reservoir temperature may be as high as 150 C. The water is characterized by a discharge temperature that reflects the permeability structure of the faults and the basement as described by theoretical models of fluid flow in fault zones (Lopez & Smith 1995). Hydrothermal circulation systems, ascent and descent channels alike, require open highly permeable steep fracture systems. Ascent channels must be extremely conducting structures permitting high flow rates in order for hot water to reach the surface environment at high tempera-

8 168 I. STOBER & K. BUCHER Fig. 3. Section from the Rhine River rift valley to the Black Forest basement along the Kinzig River valley. Horizontal scale of section about 20 km. Recharge water from the Black Forest is channeled along a fault system to the surface and does not reach the Rhine graben structure. For more details, see Stober et al. (1999). ture and appear as a hot spring. The ascent flow may follow zones of strongly fractured granite (Fig. 2) or low-permeability fault zone (Fig. 3). Driving force for the circulation is topography in both examples. In areas lacking topography, or in the absence of suitable steep highpermeability structures, deep-reaching vertical circulation systems may not develop, and natural hot springs will not be present. REACTIVE FLUID FLOW IN THE CRUST AND ITS EFFECT ON PERMEABILITY At depth below the deepest borehole that has been hydraulically tested, the permeability of the continental crystalline crust can be derived from other data and observations including geothermal data (Ingebritsen & Manning 2010). An important source of information for understanding permeability and permeability evolution with time in deeper parts of the brittle crust is the preserved effects of fossil fluid rock interaction in surface outcrops of fractured rocks from the zone between the deepest wells to the brittle ductile transition zone. The temperature range where these structures formed is about C. Fluid flow along fractures in crystalline basement rocks is generally accompanied by chemical reactions between the aqueous fluid and the rock exposed along the fracture. The reactions occur because the fluid is rarely in chemical equilibrium with the mineral assemblages of granites and gneisses predominantly present in the continental crust (Garrels & Howard 1959). At the generally relatively low temperature above the ductile brittle boundary (<400 C), reactions involving primary silicates are generally slow (Browne et al. 1989; Brantley 2004; Yadav & Chakrapani 2006) and most high-grade minerals are not stable in the presence of H 2 O at the prevailing conditions. Fluid rock reaction (retrograde metamorphism) occurs from the ductile brittle boundary to the surface in fractured basement that ranges in temperature from about 400 C down to ambient temperature. Hydration of primary high-grade assemblages is the major chemical process. Newly formed clay, zeolite, Fe-hydroxide, and other alteration products are common on fractures studied at drill cores from deep boreholes and from tunnels (Stenger 1982; Bauer 1987; Borchardt et al. 1990; Borchardt & Emmermann 1993; Lindberg & Siitari-Kauppi 1998; Iwatsuki & Yoshida 1999; Weisenberger & Bucher 2010). This shows that fluid rock interaction and associated mineral dissolution and precipitation along the fractures contribute to the variation of the permeability with time (Moore et al. 1983; Polak et al. 2003). Mineral coats along the fractures develop from the instant of fracture formation and its saturation with aqueous fluid until the fracture becomes sealed by reaction products. It is difficult to relate veins of observed secondary minerals in drill cores to the flow-active fracture porosity and to the quantitative permeability history of the crust. However, these structures relate to fossil flow and reaction systems and the observations can be regarded as analogs for present-day processes progressing in the deeper parts of the upper crust. There is abundant evidence for permeability-relevant reactions in deep fractured rocks. The effects of the reactions can be easily recognized in so-called reaction veins (Fig. 4). The dissolution and precipitation processes may proceed at different time-varying rates. Reaction veins have a structural component caused by rock deformation and a chemical reaction component that follows from irreversible reaction of the advective fluid with the exposed rock. Both components have consequences for the permeability of the fractured system and its variation with time. Simple monomineralic veins may form by a crack seal mechanism (Durney & Ramsay 1973; Ramsay 1980; Ankit et al. 2013). These veins have a predominantly deformational component and include quartz veins in quartzite (or granite) and calcite veins in limestone and marble. Reactive flow may dissolve components from the exposed rocks,

9 Hydraulic conductivity and fluid flow 169 Fig. 4. Serpentine reaction veins in peridotite (Erro Tobbio mantle, Italy). The serpentinization of peridotite occurred along symmetrical reaction veins along brittle fractures. During the active period of serpentinization, the permeability of the fractured rock permitted flow and reaction of aqueous fluid. Flow and reaction ceased because of self-sealing of the fractures, preventing the peridotite (brown crust) from being completely serpentinized. Later, the fractures became completely sealed, and the vein system does not contribute to the near-surface permeability of the partly serpentinized peridotite. Fig. 5. Talc veins in peridotite (Vlisarvatnet, Norway). Low-temperature hydration of peridotite (coarse-grained spinel harzburgite with brownish weathering rind) produced white silvery zones of talc along open brittle fractures that represent fluid-conducting structures. Note the vertical orientation of the fractures and the large aperture of these young structures. High content of dissolved silica in the hydrothermal fluid resulted in talc formation rather than serpentinization (Fig. 4). The open fracture system contributes to the present-day permeability of the rocks (and could be measured with well tests). import dissolved components from external sources, and precipitate new minerals in the fractures that were not initially present. These veins result from a fracture reaction seal mechanism (Bucher-Nurminen 1989). Examples include serpentine veins in peridotite (Fig. 4), talc veins in peridotite (Fig. 5), chlorite veins in biotite gneiss and in garnet amphibolite (Fig. 6), and tremolite veins in marble (to name a few examples). The structures shown on Figs 4 6 contain a wealth of qualitative information on the development and time dependence of permeability in the brittle crust. Some of the field-deduced conclusions are summarized in the figure captions. These examples suggest that vein growth starts with an initial brittle fracture, followed by a reaction period during which permeability increases, and then a period of decreasing permeability until the vein becomes impervious and fluid flow stops. Thus, permeability follows a time evolution similar to the porosity wave proposed for the lower crust (Connolly & Podladchikov 2007). The schematic permeability time relationship shown in Fig. 7 characterizes a single vein or a single fluid-conducting structure (e.g., shown in Fig. 6). The permeability evolution path (Fig. 7) shows the transient local effect of a single fracture on top of the original background permeability. After sealing the fracture, the system returns to the background permeability, which represents the sum of all effects of all flow-active fractures currently contributing to the permeability of the considered volume of rock. If many veins are active simultaneously, transient permeability increases and then, with reaction progress, permeability decreases. However, if vein formation is not strictly synchronous in an area, the single fractures contribute statistically to the background permeability. Outcrops on Vannøya (northern Norway) representatively show the sequence of a crack reaction seal cycle (Fig. 8). First, fluid-conducting fractures were formed in the green mafic igneous rock. The distinctive green brown reaction front resulted from reaction of a CO 2 -rich fluid with the green rock. These reactions destroyed all Fe-bearing minerals and carried away dissolved iron, leaving a brown albite carbonate rock. It is not a redox reaction front (e.g., Yamamoto et al. 2013) but rather a carbonation front (Priyatkina et al. 2011). Hydrothermal alteration reactions may, in their simplest form, be plain hydration reactions. No components dissolved in the fluid are lost or gained in such reactions. Hydration reactions consume H 2 O and cease if water supply to the reactive fracture ends. Consumption of H 2 O passively increases the total mineralization of the fluid (TDS). Because of hydrothermal alteration (retrograde metamorphism) of upper crustal rocks and the associated H 2 O consumption, deep fluids are generally highly saline (Frape & Fritz 1987; Edmunds & Savage 1991; Stober & Bucher 2005b; Bucher & Stober 2010). The salinity is typically much higher than that of seawater (up to NaCl saturation). Hydration reactions also desiccate the ductile

10 170 I. STOBER & K. BUCHER (A) (B) Fig. 6. Reaction veins in Caledonian amphibolite-grade gneiss: (A) Early chlorite vein in biotite gneiss (Hammerfest, Norway). Biotite gneiss has been chloritized along a brittle fracture that served as fluid conduit for hydrothermal fluid. The open fracture has been sealed later by epidote and became inactive. Young brittle fractures are open water-conducting structures and represent the present-day permeability of the rock. The late structures (blue arrows) control the hydraulic conductivity detected and measured by well tests. (B) Chlorite vein in garnet biotite gneiss (Torsnes, Kvaløya, Norway). Chloritization occurred along fractures by interaction of advecting hydrothermal fluid with the primary gneiss assemblage. Most of the fractures are completely sealed by solid reaction products. One of the fractures is only partly sealed. The visible fracture porosity contributes to the present-day permeability of the rocks (detectable by well tests). lower continental crust (Frost & Bucher 1994). Evidence for active desiccation of the lower crust is the presence of high-temperature halite in eclogites (Markl & Bucher 1998). Plagioclase is the most abundant mineral of the continental crust. This feldspar mineral is a mixture of a Ca component (anorthite) and a Na component (albite). At low-temperature hydrothermal conditions, Ca-bearing versions of the mineral cannot be in equilibrium with H2O, and it hydrates to clay or various zeolite minerals such as stilbite, laumontite, and others depending on the tempera- Fig. 7. Schematic graph showing the time-dependent permeability development related to a single fracture that serves as a fluid conduit for reactive fluids. This simple pattern can be modified in many ways, for example, by multiple fracturing, time-dependent aperture variations due to complex extension and shearing, and by externally controlled changes of fluid composition. In addition, the permeability of the large-scale permeability is controlled by the combined result of j t curves of a large number of fractures of different orientations. Fig. 8. Vein system in mafic rocks (dolerite on Vannøya, Norway). The structures suggest that first a fracture system opened, then advecting fluid reacted with the mafic igneous rock (green) producing an albite calcite rock (light brown) before the fluid conduit was finally sealed by brown (and white) carbonate in the central part of the veins. The structures support the proposed j t curve shown on Fig. 7. ture (Nishimoto & Yoshida 2010; Weisenberger & Bucher 2010) and pressure (depth). Low-temperature hydration of the anorthite component of plagioclase can be written as Anorthite þ 5 quartz þ 7H2 O ¼ stilbiteðcaal2 Si7 O18 7H2 OÞ ð9þ The mineral name stilbite is used here for a Na-free end member component stilbite Ca. It forms at very low temperature (<120 C). Quartz on the reactant side of Eq. 9 can be taken from the minerals exposed on the fracture surface or from dissolved SiO2aq in the fluid. In the pro

11 Hydraulic conductivity and fluid flow 171 Fig. 9. Laumontite (Ca zeolite) clogs vertical fluid-conducting fractures in gabbro (Langfjorden, Norway). The zeolite forms from hydrothermal alteration of primary labradorite (plagioclase) of the gabbro. The large volume increase in the solids in the reaction efficiently seals the vertical fracture system. The vein formation occurred at about 250 C and 10 km depth during the late Caledonian formation of the Langfjorden Repparfjorden fault system. cess, the albite component of the plagioclase recrystallizes as pure albite. Laumontite, another Ca-zeolite, forms from hydration of plagioclase at higher temperature (~ C) corresponding to a depth range of 6 10 km. Its formation reduces hydraulic conductivity and may efficiently desiccate the fracture system (Fig. 9). Many primary igneous minerals hydrate to zeolites and other hydrous minerals when the igneous high-temperature rocks pass through the brittle upper crust during exhumation before reaching the erosion surface. Their production on fractures typically involves a large volume increase relative to the original igneous rock. The process efficiently seals the water-conducting fracture, and the volume increase of the solids can accommodate significant fracture extension. The large volume increase of the solids in hydration reaction seals the fracture porosity and stops veins from growing. The lifetime of high-permeability conditions permitting fluid flow is difficult to quantify. However, the permeability of sealed veins may not be higher than that of the rock matrix. Generally, zeolites are widespread and common in fractured continental basement. They have been reported to occur on fractures and fissures encountered during road and rail tunnel construction in the Alps (Armbruster et al. 1996; Weisenberger & Bucher 2010), in deep drillholes in the continental basement (Stober & Bucher 2005b) and in drilled young granitic plutons (Yoshida et al. 2013). An example of zeolite-bearing veins is shown in Fig. 9. Pink Ca zeolite laumontite has formed along vertical fractures through massive coarse-grained gabbro of the Seiland igneous province in northern Norway. The exposed fractures are efficiently sealed by zeolite that formed according to a reaction analogous to Eq. 9 above. The presence of laumontite instead of stilbite as fracture filling (and other petrologic details) suggests that the hydration reaction occurred at a temperature of about 250 C. Thus the hydration veins in fractured basement displayed on Fig. 9 represents a situation at about 10 km depth, assuming a geothermal gradient at the time of hydrothermal vein formation of about 25 K km 1. Note also that the orientation of the fractures is nearly vertical as a result of predominance of vertical stresses over horizontal stresses at this depth (this is also the case shown in Fig. 5). The hydration reactions progress in two steps: A first step dissolves the reactive high-temperature minerals such as igneous plagioclase, pyroxene, and olivine. The dissolution step produces porosity and increases related permeability. Plagioclase conversion to albite and zeolite has produced initially up to 15 vol.% porosity in fractured granites of the Alps (Weisenberger & Bucher 2010). In a second step, the created secondary porosity is replaced by solid reaction products that eventually seal the open fractures. Fluid rock reactions in fractured granites thus first enhance permeability that originally has been created by mechanical fracturing of the granite. Increasing amounts of solid reaction products gradually decrease permeability of the porous, fractured wall rock rocks (Fig. 7). Hydration reactions consume H 2 O from the fluid and create a local pressure depression inducing of fluid flow toward the reaction site. In the Urach 3 borehole, vertical downward migration of fluid has been associated with progressing hydration reactions at depth (Stober & Bucher 2004). H 2 O consumption at depth may not be fully compensated by downward fluid flow because of low permeability. Consequently, a measurable vertical hydraulic potential gradient in the Urach 3 bore, with decreasing potential with increasing depth, is enhanced and supported by low and decreasing hydraulic conductivity with depth. Thus, downward fluid flow can be small despite favorable steep fracture orientation at depth simply due to low conductivity of the fractures. Deep fluid circulation systems may not develop in such low-permeability environments. There are few direct observations of changing permeability of crustal rocks with time. This is a consequence of the timescale of the fracture reaction seal processes in the crust. The effects cannot normally be detected in well tests. Even if well tests could be repeatedly performed in 5-kmdeep boreholes over periods of years, well detoriation and other technical effects would probably obliterate the signals from the undisturbed ground.

12 172 I. STOBER & K. BUCHER In tectonically active areas, water level measurements and other data collected after earthquakes have been used to show that active fracturing temporarily increased permeability. The healing of the created deformation structures occurred relatively rapidly, within the timescale of observation. Generally, the observed and documented permeability changes refer to mixed crust (basement and sediments) or ocean island crust, and not the inactive normal continental crust (Elkhoury et al. 2006; Kitagawa et al. 2007; Cappa et al. 2009; Xue & Cast 2013). Historical reports on the behavior of temperature and discharge of thermal waters in geothermal or mineral water spas indicate variations with time. One example is the spa Badenweiler in SW Germany. This large and luxurious spa was built and used by the Romans about 1800 years ago. The buildings were once heated by hot water using a sophisticated thermal heating system. Today, discharge and temperature of the hot springs are insufficiently low for spa operation. This indicates that the permeability structure has changed on the timescale of hundreds of years (Filgris 2001). FLUID FLOW AND PERMEABILITY STRUCTURE OF THE UPPER CRUST The Darcy flow law (Eq. 1b) implies that fluid flow ceases if the pressure gradient disappears. The major forces driving fluid flow in the crust are thermal disequilibrium and topographic relief. If the driving force ΔP (Eq. 1b) approaches zero, fluid flow stops. The fluid becomes stagnant. The stagnant fluid resides in the fracture pore space and chemically interacts with the rock matrix at a very slow rate. The fluid interacts with solids and pores by diffusion. In low-permeability near-surface rocks such as clay and shale, with permeability lower than m 2 (10 nd), fluids are considered stagnant, and the Darcy flow law cannot describe fluid transport because at very low permeability and feasible pressure gradients, the flow force relationship is not linear (Bear 1979). In fractured crystalline basement rocks with a typical permeability of m 2, fluid flow stops if the driving pressure gradient vanishes, for instance, because the topographical relief has been eroded and removed. Therefore, the fluid in the fracture pore space of basement of flat vast plains has been considered essentially stagnant. This view ignores a very efficient driving force for fluid flow, the Earth tides. Two times every day a tidal wave moves through the Earth crust. The rise and fall of the Earth surface amounts to some tens of cm (Emter et al. 1999). This is because the Earth is not a rigid body; it reacts elastically to the gravitational forces of moon and sun. The moving layers of rocks exert compressional and extensional forces on each other. The pore space of the rocks also deforms elastically in reaction to tidal forces, with the result that the fluid residing in the pore space experiences an everlasting alternation of compression and extension. The consequence of the ever-changing tidal forces can be observed as water table fluctuations in boreholes. The magnitude of the fluctuations depends on the position of the sun and the moon (e.g., Bredehoeft 1967; Evans & Wyatt 1984; Hsieh et al. 1988). Therefore, the highest amplitudes occur at full moon and new moon. In the fractured crystalline basement at the deep drilling site Urach 3, water table fluctuations related to tidal forces of up to 20 cm per day have been measured (Stober 2011; Fig. 10). It can be concluded that the observed fluctuations of the water table in the wellbore indicate the reaction of a very large volume of deep water. This conclusion is in accord with the very low fracture porosity of about / = and the very low compressibility of deep fluid c w = Pa 1. A pressure change of ΔP = 1 bar results in a relative change of the volume of water of DV w /V w = and relative to the crust of DV w /V rock = This implies that the fracture pore space is interconnected on a large scale and the crust reacts in a hydraulically coherent fashion. Long-term hydraulic tests in deep boreholes confirm that a large volume of fluid hydraulically reacts and that the fracture porosity of the basement is an interconnected network of water-conducting structures. The 1-year-long pumping test at the 4000-m-deep research drillhole (KTB- VB: see above) in SW Germany extracted a total of m 3 saline thermal fluid (Erzinger & Stober 2005) from the open hole at m at a rate of 0.5 l sec 1 and later 1.0 l sec 1 (Fig. 11). The composition of the thermal fluid remained constant during this time (Stober & Bucher 2005a). Given the porosity of 0.5%, the homogeneous volume of fluid originates from a cylinder of fractured rocks with a radius of about 100 m around the wellbore. The hydraulic signal reaches far beyond this cylindrical volume around the bore. The extracted fluid must be replenished from a much larger volume around the borehole. In regions with high topographic relief, deep flow systems develop in the crystalline basement. Temperature profiles in very deep boreholes can be used to estimate the water circulation depths. For instance, in the 5000-mdeep geothermal well, GPK-2 in Soultz-sous-F^orets near Strasbourg, France, (Genter et al. 2010) efficient water circulation reaches to 3700 m depth, which is 2300 m into the crystalline basement. Below 3700 m, the geothermal gradient is identical to the regional average (~27 K km 1 ). Above this depth, water temperatures indicate advective heat transfer by upwelling hot deep water. Finally, in the near surface, fluid temperature decreases very rapidly, resulting in a very high geothermal gradient (>100 K km 1 ). Thermal signatures of advective fluid

13 Hydraulic conductivity and fluid flow 173 Fig. 10. Water table fluctuations in the Urach 3 borehole caused by tidal forces. (A) (B) Fig. 11. Well test design and test response data from hydraulic tests in fractured continental basement in the KTB pilot hole at 4 km depth (Oberpfalz, Germany). flow may be absent in some areas because permeability of the basement is low or because pressure gradients gradually diminish (with the exception of the cyclic tidal forces). Temperature profiles of km-deep boreholes, data, and observations from hydraulic tests and tidal water level fluctuations all consistently show that fluids in the continental crust occupy an interconnected communicating pore space