Guardbridge geothermal technology demonstrator project: feasibility report

3. Geothermal Model Development

3.1 Geological model

A regional-scale geological model (1:100,000 - 1:250,000) was constructed by compiling the surface geology maps with 1) a digital surface terrain model (tiles NO41 and NO49), 2) projected faults and rock formation boundaries (horizons), 3) coded and georectified boreholes displaying the top of each formation, 4) modelled faults to a depth of 1000 m with an average plunge of 60°, and 5) georectified dip data and the 1:50,000 DiGMapGB cross sections for the British Geological Survey Sheets 41 (North Berwick) and 49 (Arbroath). For simplicity in this regional-scale model, faults with a throw <30 m and intrusions <500 m in diameter were excluded from the model.

3.1.1 Background geology

Based on cross section interpretation, the regional geology consists of strata dipping towards the SE. To the north of Guardbridge lies the thick Ochil Volcanic Formation (approximately 2000 m thick) consisting of olivine basaltic lavas and volcaniclastic rocks, offset by a series of normal faults with an average displacement of 200 m. The top of the lavas and associated volcaniclastic rocks of the Ochil Volcanic Formation are overlain by sandstones of the Scone Sandstone Formation which display an average thickness of 300 m. The Scone Sandstone Formation consists of purple-brown and purple-grey, fine- to coarse-grained, commonly cross-bedded sandstones with subsidiary siltstone, mudstone, conglomerate, sparse andesitic lava flows and some calcareous beds with concretionary limestones towards the top (Armstrong et al., 1985; Browne et al., 2002).

Overlying the Scone Sandstone Formation is the Upper Devonian Glenvale Sandstone Formation with gradational contacts consisting of brown, red, purple, yellow and cream feldspathic sandstones, commonly containing bands of red siltstone and pebbles of silty mudstone, but no siliceous pebbles (Browne et al., 2002). Honouring all available geological data, the cross sections display the Glenvale Sandstone Formation as having an average thickness of 600 m. Beneath Guardbridge, the top of the Scone Sandstone Formation is located at ~530 m. Although not exposed at the surface in northeastern Fife around Guardbridge, the highly and moderately productive Upper Devonian Knox Pulpit and Kinnesswood formations, respectively, sit stratigraphically above the Glenvale Sandstone Formation, and are assumed to be present in the subsurface around Guardbridge. Their type locality is in Glen Burn near Kinnesswood and Loch Leven (Browne et al., 2002). The highly folded and faulted rocks of the Inverclyde and Strathclyde groups overlie the Upper Devonian sedimentary rocks in the area to the south of Guardbridge.

3.1.2 Model construction

The Late Carboniferous Dura Den Fault with a normal sense of displacement (downthrown to the south) separates the Upper Devonian sedimentary and igneous rocks in the north from the Carboniferous sedimentary rocks in the south (Fig 3.1). The antithetic Maiden Rock Fault is downthrown to the north, and these two fault systems (Dura Den and Maiden Rock) form a graben-type structure. The Maiden Rock and Ceres faults form an en echelon (right stepping) structure such that the graben is rhomb-shaped in the vicinity of Guardbridge. Rocks in the hanging wall of the Dura Den and Maiden Rock faults have 'rollover' anticline and drag folds associated with them (not shown on Fig. 3.1).

The Dura Den Fault orientation drawn from all cross sections (6 sections in total and only two presented here) were collated in Move ^TM to produce the listric Dura Den Fault surface soling at a depth of ~6000 m (Fig. 3.2). The modelled Dura Den Fault is the best interpretation honouring all available geological data. The location and termination of the fault was determined from published British Geological Survey 1:50,000 surface data and no additional field evidence is included in this analysis as the exposure of the fault is very poor in northeast Fife. The actual location of the fault is not visible in the Guardbridge area, but the Guardbridge Paper Mill borehole (GR: 345010 719649) is within Upper Devonian rocks (Glenvale Sandstone Formation) and therefore the borehole lies north of the fault. The current model therefore depicts the whole of the Guardbridge site to be in the footwall of the Dura Den Fault, however, it is possible that the fault is further north within the southern end of the Guardbridge site (Fig. 1.1).

Fig. 3.1: Graphical representation of the Dura Den Fault (McCoss, 1987). The cross-section is schematic, combining geological observations from many sites to summarise the sub-surface geometries of all formations and faults. Abbreviations: c3 Upper Old Red Sandstone, dL1-3 Calciferous Sandstone Measures, dL4 Lower Limestone Group, STMS St Monance Syncline. Vertical exaggeration present.

Offset on the Dura Den Fault at Guardbridge (Section 3; Fig. 3.2) is estimated at 723 m (within the error of cross section construction), however it is evident from the cross-section analysis that the Dura Den Fault displacement decreases towards the fault tip in the northeast, and increases towards the southeast. Displacements increase from about 236 m to 1615 m over approximately 10 km.

The proximity of the Guardbridge site to the Dura Den Fault is therefore significant because of the variability of the geology on either side of the fault (Fig. 3.1, 3.2). The southern side of this fault contains younger rock sequences and therefore there is the potential to access the highly productive Kinnesswood and Knox Pulpit formations. Additionally, productivity within the Dura Den Fault zone could be good because the fault will have a zone of damage which will influence the important rock characteristics, such as porosity and permeability.

A fault "damage" zone was produced around the Dura Den Fault based on calculations presented in Childs et al. (2009) and this permits a prediction of fault zone width based on the amount of fault displacement and the fault geometry. Fault displacement (in metres) was measured within Move and fault zone thickness incorporates the fault core, the zone of most intense deformation associated with faulting, as well as the damage zone of related fracturing and brecciation of rock adjacent to the fault core. The core of the Dura Den Fault may be positioned within the limits of the modelled fault zone (Fig. 3.3). The calculated values used to produce the fault zone fault zone widths are presented in Table 1.

Fig. 3.2: Above: 1:50,000 Bedrock Polygon surface geology provided by the BGS. Guardbridge site represent by the black rectangle within the Glenvale Sandstone Formation ( GEF), north of the Dura Den Fault. Below: Cross-section interpretation. Formations: orange dashed - Ochil Volcanics ( OVF); green - Scone Formation ( SCN); pink - Glenvale Sandstone Formation ( GEF); blue - Inverclyde Group ( INV); purple - Fife Ness Formation ( FNB); yellow - Anstruther Formation ( ARBS); light mint green - Pittenweem Formation ( PMB); dark mint green - Sandy Craig Formation ( SCB); olive green - Pathhead Formation ( PDB); dark purple - Hurlet (Hur); red - Lower Limestone Formation ( LLGS); bright green dashed - Central Scotland Late Carboniferous Tholeiitic Dyke Swarm ( CSTD); green/blue intrusion - Scottish Late Carboniferous to Early Permian Plugs and Vents Suite ( SCPPV).

Fig. 3.3: Fault damage zone associated with the sections closest to Guardbridge (Sections 3 and 4). Boundary of the fault zone is represented by the dashed red lines.

3.1.3 Production of simplified 3D geological model

The final 3D Move model was created from the validated cross-sections and is presented as surfaces for the formations; these are either colour-coded horizons (the tops of formations) or faults. A 3D view of the model centred at Guardbridge and illustrating the fault zone displacement is presented in Fig. 3.4.

Table 3.1: Fault displacement and fault zone thickness. Sections 3 and 4 are presented in the text and the remaining sections are not presented as part of this report.

SECTION	DISPLACEMENT (metres)	FAULT ZONE THICKNESS (metres)
Section 1	235.9	20
Section 2	336.1	30
Section 3	723.3	200
Section 4	810.8	300
Section 5	1482.2	600
Section 6	1614.7	700

3.2 Limitations of the geological model

All interpretations behind the construction of the model are based on previously published maps and data, and relationships observed within Move ^TM. The lack of data is the foremost limitation in producing a high-resolution model of the Carboniferous-Devonian subsurface geology. This includes insufficient detail about the position of formation (horizon) boundaries at depth, the lateral changes in formation thickness, and the location and geometry of the Dura Den Fault and other less significant structures. Borehole coding and interpretation was affected by the lack of data and some poor data quality in the existing borehole information. An average uncertainty for depths of horizons is estimated as ±10 m. Available borehole data is limited to total depths of 20 m to 241 m, and there are limited boreholes >100 m depth in the Guardbridge area. Unit thicknesses are based on available map evidence.

Subsurface structural complexity at depth is very difficult to model without more field, seismic and deep borehole data, and the orientation of rocks units and fault geometry at depth are necessarily simplified. Initially, all faults were assumed to display an average plunge of 60° with an extrusion depth of 1000 m, but the Dura Den and Maiden Rock faults were subsequently remodeled based on the interpetations of McCoss (1987), including the listric geometry of the Dura Den Fault. This structure was modelled in Move using the orientation of the hanging wall horizons and the Constant Heave algorithm, which ultimately depends on accuracy of cross-section construction. Cross-section 4 was used as a proxy for fault construction. Other faults were modelled as planar structures at depth due to lack of available data, but in reality are most likely curviplanar horizons. No strike-slip component was taken into account during fault restorations (simple shear algorithm used), and no growth structure or damage structures are accounted for in the model.

Fig. 3.4: 3D view of the geological horizons (top of formations and fault surfaces) around the Guardbridge site looking northeast. Displacement of horizons (offset) on the Dura Den Fault is visualised and increases towards the southwest (out of the page). The target aquifer is between the purple and underlying green horizons.

3.3 Aquifer prospect evaluation

In order to investigate well design options, all the stratigraphic units and thicknesses were compiled, based on the available borehole data and the geological model presented in Section 3.1. As previously stated, the borehole penetration is no more than 241 m around the Guardbridge area and is ~100 m within the Guardbridge site, therefore depths and lithologies at deeper horizons are based on the extrapolation of the surface geology below Guardbridge and the existing cross-sections. All target horizon depths and thicknesses are based on the 2D and 3D regional geological model presented in the previous section and the closest and/or most detailed stratigraphy available (Browne et al., 1999; Browne et al., 2002; Shell, 2002; Walters et al., 2007; Dean et al., 2011). Permeabilities and porosities are estimated from published data of outcrop and shallow boreholes from Fife and the Midlothian areas (see sections 3.4 and 4.1).

This report presents four drilling scenarios for the Guardbridge Geothermal Energy Feasibility project. One on-site, relatively shallow, vertical borehole located on the footwall, and outside of the damage zone, of the Dura Den Fault. The first target is the Scone Sandstone Formation. The second drilling scenario is on the hanging wall of the Dura Den Fault in a location near the A91 and Edenside area; this well is also predicted to be outside the damage zone and to a depth of 1050 m. The target is the undivided Kinnesswood and Knox Pulpit formations. The third scenario is a deviated well starting at a second on-site location within the Guardbridge Energy Centre and deviating to a depth of 1100 m over 1000 m distance. The main target is the Kinnesswood and Knox Pulpit formations and the well will intercept the Dura Den Fault and damage zone for up to 400 m distance. The fourth scenario extends the previous well by drilling parallel to the dip of the Kinnesswood and Knox Pulpit formations to a depth of 1500 m over a total horizontal distance of about 2000 m.

Fig. 3.5: Location of three wells for Guardbridge Geothermal Feasibility Project. Wells GB-1 and GB-2 are within the Guardbridge site and Well ES is located offsite. Basic geology shown for context: pink is the Upper Devonian Glenvale Sandstone Formation, and yellow and red are the Carboniferous Anstruther and Fife Ness Fms. The blueline is the estimated trace of the Dura Den normal fault.

3.3.1 Scenario 1 - Well GB1

Fig. 3.6: Position of wells GB-1, GB-2 and ES relative to cross-section 3 (see Fig. 3.3-3.4). There are three well sites and one site (GB-2) has two total depth scenarios.

The approximate location of the vertical well within the Guardbridge site (Well GB-1) is next to the proposed new Library building (345030 719460). Figure 3.5 shows the location of Well GB-1 within the site and its position relative to the road network and Eden Estuary. The site is easily accessible from the entry point to the site. Cross-section 3 and the 3D geological model (Figs. 3.2 and 3.6) illustrate the subsurface structure and the depths of the main horizons; Well GB-1 is on the footwall of the Dura Den Fault (Fig. 3.6), approximately 500 m northeast of the fault (though this structure does not outcrop near Guardbridge and its surface trace is not well constrained). The stratigraphic log for Well GB-1 (Fig. 3.7) is based on the intercepted horizons in cross-sections 3 and 4 (see Fig. 3.3) and the target horizon is chosen to maximise depth in suitably sandstone-rich successions of the Scone Sandstone Formation, while avoiding the boundary with the Ochil Volcanic Formation. Further drilling beyond 700 m would penetrate into the Ochil Volcanic Formation and this unit continues to an estimated depth of 2500 m, though its stratigraphy is variable.

In modelling the 2D and 3D geological behaviour of the Guardbridge area, the fault orientation and damage zone have potential influence on the wells, and therefore the fault and damage zone have been modelled based on fault displacements of 723 metre and 810 metre for cross-sections 3 and 4, respectively (Fig. 3.3). The predictions for fault damage width in the 700 - 1000 metre depth range are 100 - 150 metre either side of the Dura Den Fault. The regional dip of the beds on the footwall and hanging wall are 15 ^o and 20 ^o, respectively, although the geometries and stratigraphy near the fault are not well constrained.

Scenario 1: Well GB-1

Coordinates: 345030 719460

Concerns:

• 12 m of boulder clay at top of sequence;
• shallow aquifer with 5 - 15 l/s potential flow rates in top 400+ m of Glenvale Sandstone Formation
• uncertain depth for boundary between Scone Sandstone Formation and Ochil Volcanic Formation
• potential for andesitic layers towards Scone-Ochil Volcanic Formation boundary (but well below main sedimentary target interval).

Uncertainties:

• exact position of the Dura Den Fault and the extent of the damage zone. Latter estimated to be 100 - 150 m either side of the fault, therefore well is likely to be outside of damage zone if fault is correctly positioned.

3.3.2 Scenario 2 - Well ES

The approximate location of the vertical well outside the Guardbridge site (Well ES) is next to the A91 in a local farmer's field. Figures 3.5 and 3.6 shows the location of Well ES and its position relative to road network, Guardbridge and Eden Estuary. The site is easily accessible from the A91. Section 3 (Fig. 3.2 - 3.3) illustrates the subsurface structure and the depth and thicknesses of the main horizons; Well ES is on the hangingwall of the Dura Den Fault, approximately 700 m southeast of the fault. The stratigraphic log for Well ES (Fig. 3.8) is based on the intercepted horizons in cross-sections 3 and 4 (Fig. 3.3) and the target is the Upper Devonian aquifer rocks of the Kinnesswood/Knox Pulpit formations. The modelled damage zone width is 100 - 150 metre either side of the fault and the regional dip of the beds on hanging wall are 20 ^o respectively, but geometries and stratigraphy near the fault are not well constrained.

Scenario 2: Well ES

Coordinates: 345770 718750

Concerns:

• ~6 m of unconsolidated sand and carse clay, with boulder clay at base;
• Anstruther Formation contains thin coals, plus organic-rich mudstones and siltstones which may contain oil (oil shales);
• Inverclyde Group rocks have thin evaporite and dolostone beds;
• Main target (Knox Pulpit Formation) has 5 - 15 l/s potential flow rates up to 400+ m depth, but uncertain at depths of 700 - 1250 m.

Uncertainties:

• exact position of the Dura Den Fault and the extent of the damage zone. Latter estimated to be 100 -150 m either side of the fault, therefore well is likely to be well outside of damage zone if fault is correctly positioned.

Fig. 3.7: Estimated stratigraphic log for Well GB-1 based on the 3D geological model and available stratigraphy.

Fig. 3.8: Estimated stratigraphic log for Well ES based on the geological model and available stratigraphy.

3.3.3 Scenario 3 - Well GB2

The approximate location of the top of the deviated well is within the Guardbridge site (Well GB-2 is 250 m south of Well GB-1). Figures 3.5 and 3.6 shows the location of Well GB-2 and its position relative to road network, Guardbridge and Eden Estuary. Cross-section 3 (Figs. 3.2 - 3.3) illustrates the subsurface structure and the depths and thicknesses of the main horizons; Well GB-2 starts on the footwall of the Dura Den Fault and at a depth of ~400 m begins to deviate through the fault zone and onto the hangingwall of the Dura Den Fault. The deviation requires the well to be parallel to a 20 ^o dip, and at a depth of ~1100 m, over a horizontal distance of 1000 m. The total estimated drilling distance is 1325 m. Given the estimated width of the damage zone and angle of deviation as the well penetrates the fault and damage zone, about 460 m of drilling is estimated to be through this zone. The stratigraphic log for Well GB-2 (Fig. 3.9) is based on the intercepted horizons in cross-sections 3 and 4 (Fig. 3.3) and the target is the Upper Devonian aquifer rocks of the Kinnesswood/Knox Pulpit Formation and Glenvale Sandstone Formation. The modelled drilling distances were calculated using three punctuated drops in drilling angle (from vertical to 60 ^o, 40 ^o, and finally 20 ^o).

Scenario 3: Well GB-2

Coordinates: 345110 719230

Concerns:

• 12 m of boulder clay at top of sequence;
• shallow aquifer with 5 - 15 l/s potential flow rates in top 400+ m of Glenvale Sandstone Formation on footwall; uncertain depth for boundary between Scone Sandstone Formation and Ochil Volcanic Formation on footwall; potential for andesitic layers towards Scone-Ochil Volcanic Formation boundary (but well below target interval);
• uncertain fracture network, cementation and mineralisation, and pressures across the damage zone and Dura Den Fault. Likely to intercept anticline and syncline structure in the Anstruther and Pittenweem formations.

Uncertainties:

• exact position of the Dura Den Fault and the extent of the damage zone. Latter estimated to be 100 - 150 m either side of the fault, but deviated drilling distance estimated at 460 m. Possible interception of multiple small faults and folds within the 460 m of drilling.

Fig. 3.9: Estimated stratigraphy for the deviated well (GB-2). Depth measurements are the drilled length of the deviated well and the pink interval represents the fault damage zone.

3.3.4 Scenario 4 - Well GB2

The deviated well will be oriented parallel to dip (20 ^o) at a vertical depth of ~1100 m and within the Glenvale Formation (Fig. 3.9). A continuation of the drilling at the same dip will permit the same geological unit to be targeted at a greater depth below the surface. An estimated drilling distance of 2495 m will target the Glenvale Sandstone Formation at ~1500 m below the surface.

Scenario 4: Well GB-2

Coordinates: 345110 719230

Concerns:

• Same concerns as scenario 3.

Uncertainties:

• Same uncertainties as scenario 3. Assumption that Glenvale Formation is of uniform thickness and that there are no further faults beyond the estimated damage zone.

3.4 Hydrogeological model

In order to estimate the groundwater flow rates and pathways for the HSA targets beneath Guardbridge, a review of the regional groundwaterflow system was undertaken. This provides a general overview of aquifer behaviour and the regional controls on sub-surface fluid flow. There are limitations to this analysis, however, because there is very little known about the deep sub-surface hydrogeology in Scotland, and therefore the geology beneath Guardbridge is insufficiently understood (i.e. aquifer thicknesses and porosity/permeabilities, behaviour of the Dura Den Fault); some of the key parameters required to model groundwater flow are unknown. The estimated hydrogeological properties are combined with the geological model developed in the previous sections to produce a conceptual and preliminary numerical groundwater flow model to test some necessary conditions for an economic HSA project at Guardbridge.

3.4.1 Regional groundwater flow system

The north-western and south-eastern margins of the Midland Valley, marked by the Highland Boundary Fault and the Southern Upland Fault respectively, are elevated with respect to the lower lying Forth-Clyde Axis. Within the regional context, they present areas of highest fluid potential and could provide the driving force for recharge and downward flow to a deep circulating regional groundwater system. If recharge of cool water occurs north of the Ochill Hills (north of Guardbridge), deep-seated flow may occur from north to south beneath the volcanic rocks with corresponding upwelling and discharge along the Forth-Clyde axis and near the coastline (Browne et al., 1987).

In the Eden River Valley of northeast Fife, the groundwater flow system appears to be dominated by recharge from valley sides. The majority of the recharge is either discharged directly as baseflow to the river or displaces shallow groundwater which is later discharged as baseflow. Most flow is believed to be transported via shallow flow paths (Ó Dochartaigh et al., 1999) in the upper 100 m of aquifers, and most of these are weathered and fractured and have little intergranular permeability; a small component of the flow may be feeding into a deeper regional groundwater system.

The groundwater contour pattern within the Eden River valley implies that there are two components to groundwater flow: one perpendicular from the valley sides towards the River Eden and one parallel to the length of the valley. However, this longitudinal groundwater flow component down the valley towards the coast appears to be very small and is likely to be at depth, away from the influence of the near-surface regime and constrained by the geological complexity of the area and the reduced permeability of aquifers at depth. It is likely to be slow and in the direction of the general regional hydraulic gradient (Ó Dochartaigh et al., 1999).

Groundwater chemistry of the shallow groundwater of Fife and Strathmore provides no evidence for deep flow paths. The waters are weakly to moderately mineralised and are almost invariably oxygenated with detectable concentrations of dissolved oxygen ( DO) and high Eh values (Ó Dochartaigh et al., 2006, Browne et al., 1987). There is no evidence from stable isotope and CFC analysis that these waters are especially old, though mixing between remnants of Pleistocene (more than 10,000 years old) waters and modern water has been proposed for other basins in Scotland (MacDonald et al., 2003) and cannot be wholly ruled out for Fife and Strathmore in the absence of radiocarbon data (Ó Dochartaigh et al., 2006).

Hence, evidence for deep groundwater circulation is, at best, inconclusive. Deep flows are probably small to moderate in volume, i.e. less than 10% of the total flow within the catchment, and limited to isolated discrete pathways along zones of tectonic weakness, such as faults (Browne et al., 1987).

3.4.2 Aquifer properties of the target formations

As stated earlier, the Upper Devonian rocks of the Midland Valley represent some of the highest productivity aquifers in the Midland Valley and these are present under the Guardbridge site. The Upper Devonian Stratheden Group aquifer incorporates, in ascending succession, the Burnside, Glenvale, Knox Pulpit and Kinnesswood formations and its base is marked by an unconformable contact with the Lower Devonian Scone Sandstone Formation (Arbuthnott-Garvock Group); the latter is transitional with the largely impermeable lavas of the Ochil Volcanic Formation (Browne et al., 1987).

The sandstones of the Stratheden Group and Scone Sandstone Formation are proven aquifers in Fife. The structure of the aquifer outcrop is largely controlled by extensional faulting, with much of the aquifer being fault-bonded by the SW- NE trending Fernie and Dura Den faults. The Scone Sandstone Formation is classed as highly productive (MacDonald et al., 2004) with normal operating yields in the Devonian sedimentary rocks in the range of 5 to 15 litres/second [l/s] (Ó Dochartaigh, 2006). Groundwater flow is dominated by fracture permeability, even in the sandstone formations where intergranular permeability is relatively high and anisotropic, suggesting that there may be preferential horizontal flow along bedding planes (Ó Dochartaigh, 2006). Measurements of the intergranular porosity and permeability are not available for the Scone Sandstone Formation, but measurements at one borehole in the Lower Devonian sedimentary rocks in the Strathmore Basin (Fig. 1.2) at depths of between 7 and 147 m below ground level indicate a median porosity of 14 % and a median hydraulic conductivity of 0.0014 m/d [metres/day] (Ó Dochartaigh, 2006), which is similar to the Upper Devonian Glenvale Sandstone Formation in Fife (Ó Dochartaigh, 2004). The transmissivity of the Arbuthnott-Garvock Group (undivided), which contains the Scone Sandstone Formation, is given by Ó Dochartaigh et al. (2006) as between 4 - 290 m ²/d with a median value of 34 m ²/d (6 samples), while specific capacity ranges between 2 and 258 m ³/d/m with a median of 25 m ³/d/m (7 samples). Storage in the Lower Devonian aquifer is given as an average value of 0.002 (5 samples).

Little is known directly about groundwater flow in the Devonian volcanic rocks, although fracture flow is likely to dominate, except along the boundaries of individual lava flows which may be preferentially weathered, increasing the local intergranular permeability. High flow rates in the Ochil Volcanic Formation occur in boreholes in Dundee. Intercalations of volcanic rocks within the Scone Sandstone Formaiton are likely to restrict groundwater flow both vertically and laterally.

The Knox Pulpit Formation, together with the overlying Kinnesswood Formation, generally has the highest porosity and permeability of the Upper Devonian of Fife. The underlying Glenvale and Burnside formations tend to have lower permeability, but provide significant yields in some cases. Public supply boreholes abstracting from the Knox Pulpit and Kinnesswood formations, such as Freuchie and Newton of Lathrisk, provide yields of up to 46 l/s, while those constructed in the Glenvale and Burnside Formations, such as Kinneston and the Kinnesswood boreholes, do not generally yield more than 28 l/s. The highest permeability in each of the Upper Devonian units tends to be in the uppermost 10 to 15 m of the saturated zone, where weathering has significantly increased secondary permeability (Foster et al., 1976). Porosity in the aquifer is generally relatively high. The sampled values range from 4 to 30%, with a geometric mean of 19%. Laboratory measurements of pore-size distribution and centrifuge specific yield for the same core samples show that the specific yields of sandstones with porosities exceeding 20% are likely to reach 12 to 15%. Sandstones with porosities of less than 20% tend to have more variable pore size distributions and may have specific yields of less than 5% (Foster et al., 1976). Hydraulic conductivities of 0.5 m/d (7 samples) are reported for the (undivided) Upper Devonian aquifers in Fife (Ó Dochartaigh et al., 2015). Transmissivity in the Knox Pulpit Formation is generally around 200 m ²/d. Very high transmissivity values in the Kinnesswood Formation (Kinnesswood borehole) may be explained by the fact that the area is highly faulted. In comparison, testing of the Kinneston borehole (Glenvale) gave a very low transmissivity of only 12 m ²/d (Ó Dochartaigh et al., 1999).

The higher permeabilities at outcrop are not representative of the deeper subsurface due to compaction and mineralisation. Groundwater flow can be dominated by fracture permeability, even in sandstone formations where intergranular permeability is relatively high. However, the majority of fracture inflows occur within 60 to 70 m of the ground surface. At greater depths, secondary voids also occur, but to a lesser extent. In the Kettlebridge borehole, for example, which is 123 m deep, only 10% of the total yield derives from below 100 m (Foster et al., 1976). Theoretically, fractures are likely to be closed (or absent) at depths of one kilometre or more beneath the central Midland Valley (Browne et al., 1987).

The permeability of the most deeply buried sandstones in Fife is estimated by Browne et al. (1987) to be of the order of 0.014 m/d perhaps attaining 0.14 m/d within selective but isolated zones, with a transmissivity of 20 m ²/d for the Knox Pulpit Formation (at 500m depth). Core samples suggest that the hydraulic properties of the target formations become less favourable with increasing depths, as mineral overgrowths and pressure solution reduces the porosity. Borehole geophysics further suggest reduced permeability with depths (as inferred from deep boreholes in the Knox Pulpit Formation (and parts of adjacent formations). However, from a comparison with 4 m ²/d at Marchwood (1666 - 1725 m) and Southampton (1729 - 1796 m), 7 m ²/d at Larne (968 - 1616 m) and >60 m ²/d at Cleethorpe (1100 -1498 m), the Upper Devonian/Lower Carboniferous aquifer in Fife could be capable of supporting the level of abstraction required for low enthalpy geothermal projects, although the abstracted fluids are likely to be mineralised. Since there is a lack of deep boreholes (> 500 - 1000 m) through these aquifers, and current measurements are derived from much shallower depths, it is not possible to predict flow rates and transmissivities with any accuracy.

3.4.3 Dura Den Fault permeability

The role of the Dura Den Fault as a pathway for deep regional groundwater flow is currently unknown. It has been proposed that fractures and faults that are oriented parallel to the maximum horizontal stress orientation (sH _max) experience the lowest normal stresses acting across them, therefore fractures will undergo the least amount of closure and will thus be the most permeable (Heffer and Lean, 1993). However, Laubach et al. (2004) observed that at depths of >3 km, open fractures were not aligned parallel to the sH _max direction. Instead, fractures whose state of stress are close to the failure criterion are more likely to be conductive because of localized failure associated with a large shear component acting along the fracture surfaces (Barton et al., 1995). These fractures are termed 'critically stressed' fractures and are oriented approximately 30º to the maximum horizontal stress (sH _max) orientation (Rogers and Evans, 2002; Rogers, 2003).

Cherubini et al. (2014) suggest that an initial characterization of hydraulic properties of faults could be achieved through an analysis of the fault positions in relation to present-day in situ stress field, as applied by Sathar et al. (2012) for Sellafield. The current stress field of Scotland is described as near east-west extension (Baptie, 2010) with a NNW trend for the maximum horizontal compressive stress (Heidbach et al., 2008). Trending approximately north-east, the Dura Den Fault is oriented ~60-70º to the maximum horizontal stress orientation, hence it may not fall into the category of 'critically stressed fractures', although this requires local analysis (Sathar et al., 2012).

3.4.4 Hydrogeological model development

A simple conceptual hydrogeological model arising from the geological model and regional review of the hydrogeology is presented in Figure 3.10. It focuses on the aquifer target depths and thicknesses which are important parameters, and incorporates the modelling of the Dura Den Fault damage zone. The aquifer properties are presented in Table 3.2 and are taken from published literature and reports, and work conducted by Town Rock Energy Ltd (see Section 4.1). Due the lack of available data on the behaviour of the aquifer at depth, the modelling explores some basic behaviour about the rates of recharge, changing hydraulic conductivity of units surrounding the aquifer, and the behaviour of the fault zones. It is not an accurate model of the rocks below Guardbridge, and the results therefore have large uncertainties, but it can be revised later if any test drilling programme goes ahead and relevant data become available.

Fig. 3.10: Conceptual hydrogeological model based on geological model development in Section 3.1 Simulations do not model the deviated well orientation and this is shown for context only.

FEFLOW ^® is a finite element model that simulates groundwater flow, as well as mass and heat transfer, through porous and fractured media (Diersch, 2005). As a geothermal heat modelling tool, FEFLOW ^® can simulate variable fluid density and heat transport, but some constraints on parameters such as porosity, permeability, aquifer thickness, sources of recharge and recharge rate are needed. For the purposes of this feasibility study, and in the absence of well constrained parameters, the simulations that have been performed test the behaviour of the aquifer under a range of reasonable conductivities (permeabilities), test the behaviour of the fault as a flow conductor or inhibitor, and test the range of conductivities required to get a well top flow rate of 5 l/s and 15 l/s flow (see section 4 for explanation of porosity, conductivity and chosen flow rates). The runs presented are from a 2D flow model, there is no heat flow modelling and only two vertical wells (GB-1 and ES) are included in the model runs; 3D modelling, heat transport and a deviated well (GB-2) can be included in the future, but it was not possible to develop these as part of this feasibility study.

The model includes the Glenvale Sandstone, Knox Pulpit, Kinnesswoood and Scone Sandstone formations as the main aquifer units and they are assumed to be confined. The Anstruther Formation and Ochil Volcanic Formation are assumed to be non-aquifers and the possible impacts of fracture-dominated flow through, and from, the Ochil Volcanics is tested. The fault has been modelled as a discrete zone which can have a different conductivity than the surrounding aquifer rocks, and the fault zone can be further divided into a core and a damage zone with two different conductivities; these allow the influence of the fault zone on flow rates to be investigated.

Two boundary conditions ( BC) were imposed: a fixed flux (Neumann) BC along the western (and parts of top/bottom) boundary representing recharge inflows, and a fixed hydraulic head (Dirichlet) BC in the eastern top corner of the model, representing the sea boundary. The 2D model is orientated at a 30 ^o angle to the fault plane. The sea boundary is located on the right-hand (eastern) side of the model, while the left-hand (western) boundary is facing inland towards the recharge area. The slice model considers the main aquifer formations and geometries, but as stated above, it should not be considered a true representation of the Guardbridge site because of the lack of data. In order to test different abstraction scenarios using the 2D slice model, the target abstraction rates (e.g. 15 l/s) had to be scaled according to the diameter of influence of the abstraction. Hence, the model cannot be used to assess the response at the well (e.g. draw down), but gives an integrated response of the aquifer area surrounding the abstraction. The results inform on the general behaviour of the aquifer and the fault, and therefore provide some useful insight into what parameters might be required for a productive and sustainable geothermal resource.

Model sensitivity simulations were conducted in steady state to test the model behavior and to select suitable parameter sets for abstraction simulations from those shown in Table 3.2. The wide variety of conductivities and transmissivities presented in Table 3.2 are from Town Rock Energy Ltd and published literature; the range of values reflects that the published literature is based on data from shallow boreholes, while the Town Rock Energy estimates are based on rock (matrix) properties at 800 - 1200 m depths, similar to the position of the target aquifers. Model parameterisation was initially based on the Town Rock Energy values, but these were then increased to test what level of fracture permeability is required to achieve the target yields. Since groundwater level data were not available for validation, a successful run was determined by keeping the hydraulic heads across the model to less than the topographic elevation of the land surface, which along the modelled slice ranges from 0 - 45 m. All the runs presented here assume that the well is pumping at 15 l/s (see Section 4 and 5 for the choice of this flow rate) over a period of 50 years.

Table 3.2: Parameters used in the FEFLOW ^® modelling and sources of values. K is permeability and T is transmissivity. Only the TRE values are used in the modelling. Sources are TRE: Town Rock Energy; Reference 1: Browne et al. (1987); Reference 2: Ó Dochartaigh et al. (2006).

		K m/d	Porosity %	T m 2/d	Oper. Yield	K m/d	Porosity %	T m 2/d	Oper. yield
	source	TRE	TRE	TRE	TRE	Ref. 1,2	Ref. 2	Ref. 2	Ref. 2
Knox Pulpit	Min	0.00134	10	0.134			4	12	4000
	Max	0.0134	14	6.7		0.06648	30	200
	Median					0.05817	19	16.62
Glenvale	Min	0.000134	8
	Max	0.00134	9	0.067					2400
	Median
Scone SS.	Min	0.000067	7	0.04				4	5
	Max	0.00134	11	0.201				290	15
	Median					0.0014	14	34
Ochil Volc.	Min				5
	Max				15

Deep borehole data arising from oil and gas exploration report permeability in mD [millidarcys], whereas hydrogeological modelling uses hydraulic conductivity in m/d [metres/day]; Table 3.3 illustrates the terminology and units used in the hydrogeological modelling and well design and performance evaluation. By necessity, the ability for a fluid to flow through rocks will be discussed as both hydraulic conductivity and permeability and where relevant, a conversion has been provided.

An initial set of runs tested the impact of different recharge rates on the model behaviour and resulting water levels. Initially, it was assumed that 20% of the overall recharge in the catchment reaches deeper formations at the base of both vertical wells (Fig. 3.11A and 3.11B), with 10% coming from the top and 10% coming from the west. A comparison set of runs simulated lower recharge to deeper levels (10% overall), which is more realistic as discussed above (Fig. 3.12A and 3.12B). The next set of runs tested what hydraulic conductivities are required to achieve the target abstraction rate of 15 l/s. As part of this, the conductivities of the Ochil Volcanic Formation underlying the aquifers were varied to test the response of the aquifer to increased fracture flow from below (Fig. 3.13), and the fault behaviour was investigated by varying the width of the fault zone and its conductivity (Fig. 3.14).

In the majority of runs conducted, the overall abstraction rate of 15 l/s is greater than the combined recharge fluxes (negative DBC values greater than positive NBC values in Figures 3.11 - 3.13). This imbalance is typically compensated for by the release of water from storage within the aquifer, which represents an overall longterm depletion of the resource. The timescale for this is dependent on the storage capacity of the aquifer which is poorly known. Without re-injecting water into the aquifer, the resource would not be sustainable over decades. Well GB-1 is less sensitive to saline intrusion (blue flowlines in Fig. 3.11 -3.14), being further from the sea boundary in the model, whereas Well ES draws from the sea boundary because of its proximity (Fig. 3.11). The amount of saline intrusion increases as the input from deep recharge decreases (Fig. 3.12).

Table 3.3: A comparison of different units and terms for the parameters used in the hydrogeological modelling and the well design and performance evaluation.

Hydrogeological modelling (section 3)	Well design and performance (section 4)
Flow rates l/s (litres/second)	Flow rates l/s (litres/second)
Porosity (%)	Porosity (%)
Hydraulic conductivity m/d (metres/day)	Permeability mD (millidarcys)
Transmissivity m 2/day (metres 2/day)	Permeability thickness mDM (millidarcy metres)

Figure 3.11B and Figure 3.13 summarise the effect of changing the conductivity of the underlying Ochil Volcanic Formation for Well ES. In all runs, the aquifer has a hydraulic conductivity of 0.8 m/d (which is much higher than the matrix permeabilities given by Town Rock Energy Ltd and hence assumes flow in active fractures); in run 14.1, 14.2 and 14.3, the Ochil Volcanic Formation has a conductivity of 0.008 m/d (Fig. 3.11B), 0.08 m/d (Fig. 3.13A), and 0.8 m/d (Fig. 3.13B), respectively. If the latter has higher conductivities approaching that of the aquifer, ingress of sea water is reduced and duration of the resource is longer. Finally, the fault zone behaviour is presented in Figure 3.14 for Well GB-1. The first run (run 13.5) includes a 50 m wide fault zone with a conductivity of 0.8 m/d (Fig. 3.14A). The second run simulates a 100 m wide fault zone with a conductivity of 0.08 m/d (Fig. 3.14B), and the third run is a 100 m wide fault zone with a conductivity of 0.8 m/d (Fig. 14C). The higher fault conductivities presents a fast pathway for water movement from deeper horizons towards the well, but the overall sustainability will depend on the volume of water available within these horizons from deep recharge routes.

In summary, based on a set of poorly constrained parameters required to model geothermal flow, the simple model presented here suggests that 15 l/s abstraction is possible given that the aquifer thicknesses used in the model are representative and assuming that there is sufficient fracture permeability to achieve the assumed hydraulic conductivities. The runs suggest that re-injection will be required to provide a sustainable resource for decades (to 50 years), but this does not address how temperature reduces with time. The current conceptual model underlying this 2D slice model requires that the Ochil Volcanic Formation is sufficiently conductive (due to fracturing) to permit deep water flows and that 10% of the overall recharge travels via deep flow pathways towards the coast. It also requires fracture permeability in the Glenvale/Knox Pulpit/Kinnesswood and Scone Sandstone formations. However, the properties of these formations and their behaviour at depths are so poorly understood, and further data are required before more robust conclusions about feasibility and sustainability can be drawn. The amount of saline intrusion modelled here is not realistic, because of the constraints and orientation of the 2D model, and further work on a 3D model with heat transfer and more constraints on the underlying parameters will significantly improve the model results.

Fig. 3.11A: Run 13.3 testing the influence of recharge rate on the aquifer behaviour at GB-1 well. Model run assumes 20% recharge to the deeper aquifer and K values of 0.8 m/d for aquifer and 0.008 m/d for other units.

Fig. 3.11B: Run 14.1 testing the influence of recharge rate on the aquifer behaviour at well ES. Model run assumes 20% recharge to the deeper aquifer and K values of 0.8 m/d for aquifer and 0.008 m/d for other units.

Fig. 3.12A: Run 15.3 testing the influence of recharge rate on the aquifer behaviour at well GB-1. Model run assumes 10% recharge to the deeper aquifer and K values of 0.8 m/d for aquifer and 0.008 m/d for other units.

Fig. 3.12B: Run 16.1 testing the influence of recharge rate on the aquifer behaviour at well ES. Model run assumes 10% recharge to the deeper aquifer and K values of 0.8 m/d for aquifer and 0.008 m/d for other units.

Fig. 3.13A: Run 14.2 testing the influence of the Ochil Volcanic Fm ( OVF) conductivity on the aquifer behaviour at well ES. K values of 0.8 m/d for aquifer and 0.08 m/d for the OVF.

Fig. 3.13B: Run 14.3 testing the influence of the Ochil Volcanic Fm ( OVF) conductivity on the aquifer behaviour at well ES. K values of 0.8 m/d for aquifer and the OVF.

Fig 3.14: Simulations of fault behaviour. A. Run 13.5 with a 50 m wide fault zone and K = 0.8 m/d. B. Run 13.6 with a 100 m wide fault and K = 0.08 m/d. C: Run 13.7 with a 100 m wide fault and K= 0.8 m/d.

3.5 Play evaluation and de-risking

As the first deep geothermal well in the Midland Valley, a Guardbridge well has significant value in addressing the geologic uncertainties and risks for the HSA play (prospects) across the region as outlined in the previous sections. A successful well with flow at economic rates would be a major boost to geothermal heat exploitation throughout the Midland Valley. A negative result will have varying impacts regionally, depending on the reason for failure.

Outwith this feasibility project, Town Rock Energy ( TRE) have produced regional Common Risk Segment maps for HSA targets in the Midland Valley; example maps are provided in Figures 3.15 - 3.17. The study covers an area from Arbroath and the east coast of Fife to Stirling and Motherwell in the west and North Berwick in the southeast. Publicly available well data, including wireline logs and core, have been used to evaluate porosity and permeability trends with depth. Previous studies on geothermal gradient (Gillespie et al., 2013) have been verified for these wells and a temperature estimate with depth has been calculated. Gross depositional environment maps have been made by University of St Andrews for the TRE project, based on published research. This integrated risk mapping project aims to predict areas where rock type, permeability and temperature align to give favourable conditions for warm water flow from aquifers. These are very much regional maps, and any one geothermal prospect will carry local risks and uncertainties which can be investigated with a variety of geologic and geophysical techniques.

The sparsity of borehole data, and of good quality stratigraphic and sedimentological logs and core data, means that there is significant uncertainty in the Common Risk Segment Maps. For example, the primary target intervals, the Knox Pulpit and Kinnesswood formations, are found at outcrop and in a couple of deep wells, but there are no aquifer quality data for these intervals in the shallower subsurface (i.e. less than 1500 m depth). To overcome this, how porosity changes with depth has been averaged across all Carboniferous strata penetrated in wells after detailed analysis showed that this was a reasonable reduction of the data. Data from Carboniferous successions are used to model Devonian rock characteristics because those are the only available data in this region of the Midland Valley.

Fig. 3.15: Matrix permeability quality for sandstones in Kinnesswood and Knox Pulpit formations with depth to horizon as the primary control. Permeability and porosity predictions based on core and wireline data.

Fig. 3.16: Combined permeability and depositional environment map of the Knox Pulpit Fm. Green segments represent favourable characteristics due to shallow depths of burial. Red areas represent non-deposition or erosion, unfavourable facies, or poor permeability due to depth of burial.

The most critical factor in proving the HSA play in the Midland Valley is to demonstrate that economic flow rates can be achieved from the aquifers. Flow will be determined by matrix permeability, and any increased permeability associated with natural faults and fractures in the aquifer. Future wells can be optimally designed to exploit these areas of increased flow. At present, deeper targets with temperatures in excess of 60 ^oC are likely to have very poor matrix permeability based on the available data, and therefore uneconomic flow rates.

New rock properties data from a well at Guardbridge, designed to target hot sedimentary aquifers at relatively shallow depths and temperatures of 25 ^oC to 45 ^oC, will provide valuable tests of the assumptions that have been made with regard porosity and permeability depth trends. A positive result with significant flow rates of water from a defined interval at Guardbridge will provide critical datasets on flow and rock properties that are regionally transferable, and will significantly reduce the risks of exploration within the Central Belt of Scotland.

The following sections look at the specific regional impact of each of the three well targets.

Fig. 3.17: Combined map of estimated aquifer quality (see Fig. 3.5) and predicted temperature.

3.6 Regional impact of Guardbridge wells

3.6.1 Well GB1 - vertical well on site

This well is located within the Guardbridge site and targets the upper units of the Glenvale Sandstone and the Scone Sandstone formations (Fig. 3.7). Drilling this well would increase our knowledge and understanding of the local and regional stratigraphy by providing lithology and formation thickness data which could be used to improve the regional mapping of the Scone Sandstone Formation in the northeast Midland Valley and update the Devonian HSA play. This play has not been mapped in detail at this time, though initial outcrop studies have been conducted and some surface porosity and thermal conductivity data have been collected by Town Rock Energy Ltd and the University of St Andrews.

The recovery of subsurface core would permit measurements of porosity and permeability to be made on fresh rock at a depth of around 500 m. These would be rare samples and would allow comparison with outcrop porosity and permeability data from locations in Fife, Perthshire, Tayside and Angus which may have been impacted by weathering and are generally less cemented and compacted. Identifying good porosity in the Scone Sandstone Formation would be encouraging for development of the play in, at least, the central and eastern Midland Valley.

Core samples would also allow identification and measurement of natural fractures, and whether the fractures are open or mineralised. A flow test would address whether the Scone Sandstone Formation can give economic flow rates of warm water. Devonian rocks that lie beneath Fife south of the Ochil Fault have generally been assumed to be too deep and too tight to produce water without stimulation. A successful flow from the Scone Sandstone Formation would trigger a review of where this Devonian target might be present at depth in Fife.

GB-1 also targets the Ochil Volcanic Formation underlying the Scone Sandstone Formation. Similar volcanic rocks provide potable water from wells in and around Dundee at potentially economic flow rates. A demonstration of significant water flow from the Ochil Volcanics with core data that showed open fractures would be encouraging for the development of the play. This would trigger a review of the play in the Stirling, Perth and Tayside area, especially where heat demand is high (e.g. Dundee, Perth and Stirling).

Bottom hole temperature data will be valuable in determining the geothermal gradient at the Guardbridge site, and establishing whether this is on trend with regional data or whether local variations are significant.

3.6.2 Well ES - vertical well off site

Well ES is located 1 km to the southeast of the Guardbridge site and targets the Upper Devonian Kinnesswood and Knox Pulpit formations (Fig. 3.8). Figure 3.18 provides an estimate of the extent of the Knox Pulpit Formation based on gross depositional environment maps (Robinson, unpublished data) in order to demonstrate the significance and regional impact of the geothermal project at Guardbridge. The Kinnesswood, and particularly, the Knox Pulpit formations have been identified as the primary HSA targets in Fife, and towards the south and west within the Midland Valley (Browne et al., 1987; Galbraith et al., 2013). Well ES1 also targets the Devonian Glenvale Sandstone and Scone Sandstone formations which have not been mapped regionally as HSA targets due to their great depth in most areas.

The Kinnesswood and Knox Pulpit formations CRS mapping relies heavily on the porosity depth trends and porosity-permeability cross plots from TRE's regional well study. Actual data from the two formations is sparse and the opportunity to acquire data at shallower depths will reduce the current level of uncertainties. The Guardbridge site is in a low risk, green, segment of the Kinnesswood and Knox Pulpit CRS maps (Figs. 3.26 -3.17) and so a positive result will support drilling in other green segments around the region.

Fig. 3.18: Gross Depositional Environment ( GDE) map for the Knox Pulpit Formation showing an estimate of the regional extent of the aeolian (and fluvial) deposit.

A successful result which demonstrates that there is a higher overall porosity for these formations at shallow depths than has been predicted, based on trends in other Carboniferous strata, would be significant regionally. Such a result may allow the green segment on the Kinnesswood and Knox Pulpit CRS maps to expand into the yellow segments which are currently downgraded due to loss of porosity and permeability at depth. Deeper, warmer water may be more productive in terms of flow rates than currently predicted.

Success in the older Scone Sandstone Formation and Glenvale Sandstone Formation would have similar impact to a positive result at GB-1; that is, encouraging a review of this geologic interval in the area of the Strathmore Basin to the north (Fig 1.2).

3.6.3 Well GB-2 - deviated well on site

This more complex well trajectory is a hybrid of the GB-1 and ES wells. The well head is located within the Guardbridge site and the well deviates so that the bottom of the well is located about 1 km to the southeast. The primary target is the undivided Knox Pulpit and Kinnesswood formations (in the vicinity of the Dura Den Fault (Fig. 3.9). This potentially combines the highest porosity and permeability aquifer with an area of natural fractures associated with the Dura Den Fault. In an optimal scenario, flow will be enhanced by a combination of open fractures and good aquifer permeability, and more of the aquifer will be accessed by drilling at a high angle to the formation; flow along the fault zone may increase the sustainability of the aquifer. However, the fault-related fractures could be cemented and that could result in reduced aquifer permeability. A positive outcome demonstrating good flow rates in the Knox Pulpit/Kinnesswood aquifer will have the same regional impacts as Well ES. In addition, demonstration of increased flow associated with the fault would trigger a review of fault and fracture distribution and stress history in the region in order to identify other areas where faulting may enhance productivity. Success in a deviated well would also inform future well design options. Optimising well design to achieve higher flow rates will be key to achieving economic feasibility for HSA wells.