The role of stress history on the flow of fluids through fractures

Safe management and disposal of radioactive waste in an isolated sub-surface setting is necessary to minimize the long term effects of radioactive waste on the biosphere. Deep geological disposal is currently the only available pragmatic solution for the storage and safe disposal of radioactive waste in Europe. According to the current concepts, high level and long lived radioactive waste and spent fuel are intended to be stored/disposed of in a geological disposal facility (GDF) within a stable geological formation (host rock) at depth (typically ~50–800 m) beneath the ground surface. Hence, the radioactive waste is securely isolated and contained. At depth the rock mass may be a naturally fractured environment and the excavation of underground repositories is recognized to induce additional fractures (Bossart et al., 2002; Rutqvist et al., 2009). The zone of the host rock in the vicinity of the underground opening whose hydromechanical properties are modified as a consequence of excavation activities are referred to as excavation damaged zones (EDZ) (Tsang et al., 2005). Depending on the in situ stress conditions, preferential pathways may form along faults, joints and the EDZ.

Fluids, such as gases and water, are expected to play a role in the transport of radionuclides away from the GDF. The conductivity of fluids through a fault or fracture is understood to be controlled by the interplay of fracture orientation and principal stress directions (Barton et al., 1995; Finkbeiner et al., 1997). Around the GDF there are two distinct zones with differing fracture orientations, fracture densities and fluid flow properties: (1) the EDZ where an intricate range of fracture orientations are present in a complex localized-stress field; and (2) the far field zone where pre-fractured host rock may be present.

It has been proposed that fractures and faults that are oriented parallel to the maximum horizontal stress orientation (σ_Hmax) experience the lowest normal stresses acting across them and therefore will undergo the least amount of closure and will thus be the most permeable (Heffer and Lean, 1993). This is based on the assumption that the fractures and faults experiencing the least amount of stresses will offer minimum resistance to flow and therefore will have relatively high permeability. However, observations by Laubach et al. (2004) on a number of basins in western United States using core permeability, stress measurements and fluid flow datasets have shown that at a depth of >3 km, the open fractures are not aligned parallel to the σ_Hmax direction as previously understood. Hence, in situ stress orientations cannot be realistically used as an indicator for predicting fluid flow in fractured rocks.

When the shear stress and normal stress on the faults and fracture surfaces are plotted with respect to the in situ stress field in a Mohr space, the faults and fractures that are scattered above the Mohr-Coulomb failure criterion are termed critically stressed and hence expected to be conductive (Fig. 1). Critically stressed fractures are expected to be present amongst the numerous complex fracture networks within and around the EDZ and in the far field around the GDF.

In the present integrated field and experimental study, the relationship between fracture and stress tensor orientations with respect to the flow of fluids through fractures in Sellafield area, UK, has been investigated. A reappraisal of the complete dataset has been undertaken (Reeves, 2002; Reeves et al., 2003) in order to apply the critical stress approach to the Sellafield area of the UK. The distributions of conductive and non-conductive fractures with respect to in situ stress conditions are examined to comprehend the relationship between in situ stress and fracture flow. Furthermore, the effect of loading and unloading on the flow of fluids (water and gas) through an idealized experimental fracture surface and its implications for fluid flow have also been investigated.

The Sellafield area, in west Cumbria lies in a transitional structural zone between the western margin of the Lower Palaeozoic Caledonian basement (Lake District massif) and rocks of the younger East Irish Sea Basin (Michie, 1996). The lower Palaeozoic basement rocks consist of a thick pile of extrusive, volcanic and volcaniclastic rocks known as the Borrowdale Volcanic Group (BVG). Sandstones, mudstones, and breccias of Permo–Triassic age make up the rocks of the East Irish Sea Basin.

As a part of the UK Nuclear Industry Radioactive Waste Executive's (Nirex) Sellafield investigations, 29 deep boreholes were drilled totalling about 27,500 meters in overall length (Bowden et al., 1998). Systematic geological logging of the cores was performed followed by mineralogical, geochemical and geotechnical studies. In situ stress analysis was carried out to understand the present day stress conditions of the rocks (Nirex, 1996). In situ hydraulic testing was also performed to assess the hydraulic properties of the rock mass to identify the flow zones (Armitage, 1996).

The BVG was planned as a possible host rock for the future radioactive waste repository in the Sellafield area. The BVG consists of Ordovician (Caradocian) pyroclastic volcanic rocks which originated from a subaerial, subduction related, calc-alkaline volcanic field that resulted from the closure of the Iapetus ocean (Millward et al., 1994). The matrix permeability of fluids through the BVG is very low. However, fluid flow through BVG has been determined to occur via complex networks of discontinuities such as faults and joints existing within them (Nirex, 1997).

The dataset presented here was acquired by Nirex as a part of their preliminary investigations into the Sellafield area to assess its suitability as a potential site for deep underground disposal of intermediate-level and long-lived low-level radioactive waste. Fracture orientation datasets, in situ stress measurement datasets, and fluid flow indicators (from hydrogeological production tests, short interval tests and mineralogical and/or petrographic examination of discontinuities within the borehole core) were collected during the drilling and geophysical logging of the 29 deep boreholes scattered around the Sellafield area (Armitage et al., 1996; Bowden et al., 1998; Nirex, 1996; Nirex, 1997).

Ultrasonic geophysical logging tools such as borehole televiewers (BHTV/UBI) were used to gather a high resolution detailed 360° record of the borehole elongation and distribution of natural fractures. Features such as breakouts were easily discerned by the prominent distortion to the waveforms in the BHTV image log as a consequence of an increase in the number of fractures encountered around the breakout (Rider, 1986; Paillet and Kim, 1987). From the sine wave fitting routines of the acoustic image from the BHTV/UBI tool, the azimuth and dip of the fractures were deduced (Reeves et al., 2002). The presence of mineralized zones in the fractures were inferred by a comparison between the amplitude of the reflected signal (AMWF) and the two way travel time (TTWF) images; with open fractures showing up in both images and mineralized fractures evident only in the AMWF image. The conductive and resistive features were identified in the borehole by means of electrical imaging tools namely, the Formation MicroScanner (FMS) and the Fullbore Formation MicroImager (FMI) (Standen, 1991). The FMS and FMI measure the ease with which electrical current can be conducted by the borehole walls and aid in inferring zones of high conductivities which are marked by the presence of pore fluids within fractures or flow paths. The FMS/FMI logs display sharp electrical contrasts in the borehole images depending on the presence of conductive fluids, cemented horizons, and also helps to decipher geometrical relationship such as high dip in a sequence with low structural dip (Rider, 1986).

Regional stress maps for the Sellafield area were produced by Brereton et al. (1997) from the information gained from the breakouts observed within the BVG and the Permo–Triassic rocks within the locality. From the analysis of elongation of the borehole cross-section, in situ stress orientation data was acquired.

Figure 2 shows the UK regional stress regime (Evans and Brereton, 1990) displaying the distinctive NW–SE (135/315°) σ_Hmax in the UK and exhibits the location and alignment of the σ_Hmax oriented breakout results derived from the in situ stress investigations of the top hundred metres of the BVG in the Sellafield area in addition to its relationship to the existing geological structures. The data demonstrates a mean orientation of 159/339° for the σ_Hmax orientation with values ranging from 118/298° to 190/010° and came from the reappraisal of the stress tensor orientations undertaken by Reeves (2002).

A frequency plot of the orientations of fractures observed within the flowing and non-flowing area identified from the short interval test (SIT) investigation from Sellafield Borehole RCF3 (Armitage et al., 1996) shows that the flowing fractures are not oriented parallel to the σ_Hmax direction (Fig. 3). On the contrary, non-flowing fractures were observed to be aligned parallel to σ_Hmax direction. From the critical stress analysis of the flow zones and the short interval test zones in Sellafield Borehole RCF3, no critically stressed fractures which are in equilibrium with the in situ stress regime, could be established (Fig. 4). All the flowing and non-flowing fracture data points plot below the Mohr-Coulomb failure envelope for a coefficient of friction of 0.6. Moreover, the flowing and non-flowing fracture datasets are dispersed across the Mohr diagrams with the flowing fractures forming a subset in the upper zone of the Mohr space whereas the non-flowing fractures are randomly scatter around the lower zones within the Mohr space. These observation would suggest that the critical state approach is not applicable to this example in the Sellafield area, which is also concluded from the in depth study presented in Reeves (2002).

All of the experiments reported in the present study were performed using the bespoke angled shear rig (ASR) (Fig. 5) built at the Transport Properties Research Laboratory (TPRL) of the British Geological Survey. The ASR consists of two stainless steel blocks with a 60 mm × 60 mm × 50 μm clay gouge sample sandwiched between them. Saturated kaolinite clay paste (gravimetric water content of 80%) was used as the sample material in this study. Unlike standard direct shear experiments, the top and bottom blocks can be oriented at different angles to the normal load/horizontal displacement. Experiments have been conducted on horizontal blocks and at angles of 15, 30 and 45°. However, the current data are all for a fracture oriented at 30° to the direction of horizontal movement (or 60° to the normal stress direction). Normal load was applied using a servo controlled Teledyne ISCO-260D syringe pump pressurizing an Enerpac single acting hydraulic ram connected to a rigid loading frame. Pore fluid (water or helium) was introduced through the centre of the top block by means of a Teledyne ISCO 500D syringe pump connected in series to a gas-water interface vessel. Pore pressure transducers, attached to ports which are positioned orthogonally to each other at 15 mm from the central pore fluid inlet allowed measurement of pore pressures within the fracture surface (Fig. 5a).

The complete experimental programme conducted in excess of 40 separate experiments. The results from four experiments are presented and summarized in Table 1. These experiments were conducted at constant injection pressure with cyclic normal load steps from 0.3 to 3 to 0.3 MPa in a total of approximately 25 steps. Repeat results are presented for two water injection and two gas injection tests. Note that although the ASR is capable of performing tests during active shear, the presented results were conducted without shear. However, the slope of the experimental fracture plane resulted in an unavoidable shear movement of ~5 2 μm during loading from 0 to 3 MPa. The evolution of shear load with loading was measured using a load cell fitted laterally to the top-block (Fig. 5).

A pore water pressure of 1 MPa was used during water injection tests, which was of sufficient magnitude to initiate flow. A higher gas injection pressure of 3.55 MPa for ASR_Tau07 and 4.95 MPa for ASR_Tau08 was necessary in order to initiate flow. Normal load on the fracture surface was varied approximately every 24 h to give optimum time for attaining stress and pore-water pressure equilibrium within the fracture surface. Data were logged every 2 minutes using LabVIEW. The data were processed based on the formulations detailed in Gutierrez et al. (2000) and Cuss et al. (2011).

The results of water and helium injection experiments during a loading and unloading cycle are presented in Figs 6 and 7, respectively.

As seen in Fig. 6, water flow rate decreased during loading by approximately an order of magnitude, but only partially recovered during unloading (hysteresis). In test ASR_Tau01 flow rate decreased from 86 to 6 μl h⁻¹ during loading from 0.2 to 3 MPa. During unloading from 3 to 0.1 MPa flow rate recovered from 6 to 23μl h⁻¹ (Fig. 6b,c). The shear stress to normal stress ratio decreased from a starting value of 1.3 to 0.7 during loading, whereas during the unloading the shear stress to normal stress ratio increased steadily with each unit step in the unloading cycle to a final value of 2.3 at 0.19 MPa (Fig. 6d). Similar to ASR_Tau01, in test ASR_Tau05 flow rate decreased from 48 to 5 μl h⁻¹ during loading from 0.2 to 3 MPa and the flow rate recovered to 10 μl h⁻¹ during unloading to 0.1 MPa (Fig. 6f,g). Shear stress to normal stress ratio decreased from 1.5 to 0.7 during loading and during unloading the ratio increased steadily with each step to finally achieve 2 at 0.22 MPa (Fig. 6h). Even though the initial flow rates in ASR_Tau01 and ASR_Tau05 were remarkably different (86 and 48 μl h⁻¹, respectively) prior to loading, the flow rate at maximum loading is similar for both tests, with values of 6 and 5 μl h⁻¹, respectively (Fig. 6c,g). Furthermore, the shear stress to normal stress ratio evolved in a similar way in response to normal load variations in both tests (Fig. 6d,h).

As seen in Fig. 7, helium flow rate evolution during cyclic loading was remarkably different to that observed during water injection. In ASR_Tau07, gas flow rate increased and decreased arbitrarily during loading from 0.25 to 1.5 MPa. However, during further loading from 1.5 to 3 MPa a reduction in flow rate with increasing load was observed; flow rate decreased from 200 to 90 μl h⁻¹ (Fig. 7b,c). In contrast to unloading during water injection, gas flow rate increased by an order of magnitude from 90 μl h⁻¹ to in excess of 1000 μl h⁻¹ during unloading from 3 to 0.25 MPa (Fig. 7c). The flow rate was observed to rise when the shear stress to normal stress ratio was greater than unity during the unloading cycle (Fig. 7d). Similar complexity in flow rate evolution during loading and unloading was observed in test ASR_Tau08. The gas flow rate increased and decreased during loading between 0.4 and 1.2 MPa. Loading stages from 1.2 to 3 MPa resulted in a steady decrease in flow rate from 550 to 276 μl h⁻¹ (Fig. 7f,g). The shear stress to normal stress ratio during loading and unloading was similar to that observed in ASR_Tau07 (Fig. 7d,h). Significant gas flow rate increase occurred when the shear stress to normal stress ratio increased above unity during unloading.

A comparison of the results from the water and gas injection experiments reveals that flow rate generally decreased during loading and increased during unloading. Hysteresis in flow was observed during a loading–unloading cycle in both the water and gas injection experiments. The responses of water and gas flow were remarkably different during the unloading stages. Shear stress versus normal stress plots for both water and gas injection experiments showed similar behaviour with a prominent hysteresis in shear stress during the unloading stage (Fig. 8).

It has been reported in the literature (Heffer and Lean, 1993) that fractures and faults oriented parallel to σ_Hmax direction are likely to be most conductive. However, field observations from the Sellafield area suggest that there is no apparent relationship between the fracture orientation and the σ_Hmax direction. Additionally, open and flowing fractures were not observed to be oriented parallel to σ_Hmax direction (Fig. 3). The present observation is in agreement with the misalignment of σ_Hmax and open fractures in the East Texas Basin, Val Verde Basin, Green River Basin and Powder River Basin reported by Laubach et al. (2004). Interestingly, in the present study, non-flowing fractures were observed to be oriented relatively parallel to σ_Hmax direction affirming that σ_Hmax direction on its own accord cannot be used as a predictive method for fluid flow directions as proposed previously. Furthermore, critical stress analyses of the rocks in the Sellafield area revealed that no fractures were critically stressed and are in frictional equilibrium with the in situ stress field.

One possible explanation for the non-applicability of critical stress analysis to the Sellafield area is the complex tectonic history. Lewis et al. (1992) have proposed a rapid Palaeogene–Neogene uplift and inversion, with uplift of approximately 2 km (Chadwick et al., 1994). This has resulted in stress relief along fractures and faults to the present stress state which is below critical stress conditions.

The experimental results confirm a strong relationship between normal stress and flow, as expected. Water and gas flows were significantly decreased due to increasing normal stress. The decrease in flow is mainly due to the constriction of fluid pathways due to a decrease in fracture width during normal loading. Similar decrease in fault flow with increased effective normal stress has been reported by Brace et al. (1968), Huenges and Will (1989), Pratt et al. (1977) and Rice (1992). The arbitrary response of gas flow during the initial loading stages might have been due to the formation or enlargement of gas pathways as a result of the movement of the fracture blocks under low normal stress during initial loading. However, gas flow was significantly enhanced during unloading which suggests that water and gas flow responds in a significantly different way. The desiccation of the fault gouge by the gas resulting in enhanced flow rates cannot be entirely ruled out. However, the rapid response of flow rate to variations in normal stress during unloading suggests that desiccation would not have played a major role in significantly affecting the flow rate.

Zoback et al. (1985) and Brudy et al. (1997) have shown that the ratio of shear stress to normal stress is crucial in controlling permeability and in the movement of gas through fractures. The close relationship between fracture flow and the shear stress to normal stress ratio during the unloading stages in the present experiments also points towards its significance in understanding the flow of fluid through fractures in the BVG within the Sellafield area that have undergone stress relaxation via uplift and inversion since the end of Cretaceous. Moreover, experimental results have shown remarkable hysteresis in the shear stress versus normal stress plots (Figs 6 and 7). The hysteresis in shear stress during unloading may be attributed to the cohesive strength of the kaolinite clay gouge. In the case of a fractured rock formation undergoing considerable uplift, such as the BVG in the Sellafield area, stress relaxation is likely to result in high shear stress to normal stress ratio. Fluid flow under high shear stress to normal stress ratio as observed in the experimental fracture study becomes very probable and may explain the occurrence of flowing fractures in the Sellafield area. This highlights that an understanding of the stress history of the fracture or fault surface is essential to effectively predict the present fluid flow properties of those features. A mere knowledge of the present day in situ stress is insufficient to describe the flow properties of subsurface fluids through fractured rocks. The prevalence of flowing fractures in the Sellafield area irrespective of the current stress state not achieving critical values implies that the once critically stressed rocks have undergone stress relaxation during the Palaeogene–Neogene uplift.

The fluid flow properties of fractures within the BVG in the Sellafield area have been investigated in the present study. The results show that the stress state is below the critical stress criterion and indicates that the rocks have undergone stress relaxation as a result of exhumation associated with the complex geological history of the area. Experimental fracture flow studies performed to understand the occurrence of flowing fractures have shown that:

1. Fluid flow through fractures decreases with increase in normal stress, as expected.

2. The effect of stress relaxation on water and gas flow through a fracture surface is remarkably different. During water flow experiments considerable hysteresis is seen with only a partial recovery of the initial flow rate. Hysteresis is also seen in gas flow experiments. However, flow became much greater than the initial flow.

3. Understanding the shear stress to normal stress ratio is vital in predicting the flow properties of fractures and faults. Features experiencing high shear stress to normal stress ratios are expected to be more conductive. Moreover, high shear stress to normal stress scenarios are likely to be more prevalent in regions experiencing stress relaxation due to structural uplift or removal of the overburden as in the case of the Sellafield area.

4. Substantial hysteresis in shear stress during unloading (stress relaxation) indicates that fracture flow is significantly altered by the stress history.

There is considerable field evidence of the applicability of the critical stress theory to numerous basins worldwide. It had been anticipated that this approach would aid the understanding of the hydrogeology of the Sellafield area and it was unexpected that the theory was not supported by the field data. It can therefore be concluded that uplift and a complex stress history has a profound impact on the critical stress theory and the flow properties of faults. The stress history has to be included in an analysis and this hypothesis is supported by the experimental observations where considerable hysteresis is observed in flow and stress during unloading.

A part of the research presented here was funded by the European Atomic Energy Community's Seventh Framework Programme (FP7/2007-2011) FORGE project under Grant Agreement No. 230357. This paper is published with the permission of the Executive Director, British Geological Survey (NERC).