GIS analysis for the selection of optimal sites for mine water geothermal energy application: a case study of Scotland's mining regions

The study area spans the principal coalfields of the Midland Valley of Scotland (MVS). This is a large graben-like structure, bounded to the north by the Highland Boundary Fault and to the south by the Southern Upland Fault (Cameron and Stephenson 1985). It contains the cities of Edinburgh and Glasgow, and also the catchments of the rivers Clyde (flowing west) and Forth (flowing east). The MVS preserves a thick sequence of post-Caledonian Orogeny sedimentary rocks of Devonian and Carboniferous age, together with volcanic lavas and intrusive dolerite sills of similar age. The Carboniferous of the MVS can be subdivided (Monaghan 2014) into:

• the Scottish Coal Measures Group (Westphalian age);

• the Clackmannan Group (Namurian and Visean age), which can further be subdivided into the Passage Formation, the Upper Limestone Formation, Limestone Coal Formation and the Lower Limestone Formation – these comprise deposits of shelf carbonate, and fluviodeltaic and deltaic facies;

• the Strathclyde Group (Visean age), which hosts the West Lothian Oil Shale Formation; and

• the Inverclyde Group (Visean).

The majority of the workable coal seams are hosted by the fluviodeltaic sediments of the Westphalian Scottish Coal Measures and the Namurian Limestone Coal Formation. Laterally, the worked coal deposits can be subdivided into the Central, Ayrshire, Lothian, Fife, Sanquhar and Douglas coalfields (Fig. 1). The coals have been worked for many centuries but especially in the nineteenth and twentieth centuries. The last mines to be closed were Frances (Fife, in 1995), Monktonhall (near Edinburgh, in 1998) and the very extensive Longannet complex (Fife, in 2002). Following abandonment, dewatering pumping typically ceased and mine water heads recovered over the course of the subsequent years. Rising mine water levels have often been controlled by continued pumping (and water treatment at surface) on the part of TCA to prevent uncontrolled surface breakout (e.g. at Frances, Blindwells and Polkemmet: Chen et al. 1999; Nuttall and Younger 2004; Younger 2012; Wyatt et al. 2014; Zebec Biogas 2022). In some coalfields, the process of rebound is still underway (e.g. the Midlothian coalfield, where mine water is breaking out near Dalkeith: Jackson 2022).

Scottish mine water blocks were found to have mean geothermal gradients of 29.8°C km⁻¹ in the Central Coalfield, 26.8°C km⁻¹ in Ayrshire, 24.2°C km⁻¹ in Lothian, 22.2°C km⁻¹ in Douglas and 21.9°C km⁻¹ in Fife (Farr et al. 2021).

In this study, six datasets were compiled, as summarized in Table 1 and discussed below. Datasets 1–4, on the geometry of mine workings and monitored mine water head, were obtained from TCA.

The ‘Underground working’ vector dataset consists of 2D polygons that represent the geographical extent of underground mine workings (mostly coal but the dataset also includes some Carboniferous limestone, oil-shale or ironstone workings associated with the coal-bearing strata), georeferenced and digitized from the comprehensive collection of mine abandonment plans hosted by TCA. The dataset was originally created by TCA for automated provision of coal mining reports on ground stability and potential mining hazards (Tipper 2015b). The geographical accuracy of this dataset will be affected by human error during the original surveying (which is likely to decrease with time as surveying methods became more standardized and accurate), and possible inaccuracies introduced when georeferencing the paper plans for digitization. It is also accepted that not all mines in Scotland are recorded: the age of the first workings (twelfth century) greatly predates legislation to ensure documentation (1870s: Younger and Adams 1999), leaving some shallow mines undocumented. Mine abandonment plans in the UK became more reliable and of uniform quality after nationalization in 1946. Whilst the polygons in this dataset define the areal extent of mined coal seams, they do not show any detail concerning the layout of shafts, roadways or individual worked panels, which may influence preferred locations for drilling and accessing mine water. Moreover, they do not distinguish between collapsed longwall panels and pillar and room workings, the latter being more likely to be hydraulically open. This dataset does not contain explicit elevation data (elevation data are contained in Dataset 3).

The ‘Shallow working’ vector dataset consists of 2D polygons that represent worked portions of mined seams within 30 m of the surface. TCA created this derived dataset by extraction from the ‘Underground working’ shapefiles and keeping only portions of polygons that were within 30 m of the surface (although the surface model from which the depths were extracted is not recorded). The uncertainties inherent in the ‘underground working’ shapefiles are carried over, and many old shallow workings are likely to be absent from the dataset as they predate mandatory documentation.

This dataset does not contain explicit elevation data, other than the fact that they are less than 30 m from the surface (elevation data are contained in Dataset 3).

The ‘In-seam level’ dataset comprises of a series of points in longitude (X)–latitude (Y) space, each associated with the elevation (Z) of the seam, relative to Ordnance Datum (OD, or mean sea level). These spot elevations are digitized directly from original abandonment plans, and the spatial accuracy is subject to the same challenges as the previous two datasets (surveying and georeferencing errors), as well as potential typographical errors and errors in conversion to metric units. The information in the dataset is not uniformly distributed, and some areas have sparse ‘In-seam level’ points.

The mine water head (i.e. piezometric head within mine void aquifers) was obtained for each of TCA's monitoring stations (typically shafts or boreholes; n = 48) in the MVS. These are point data (X, Y) with associated elevation (Z) values representing the water head in metres relative to OD. The georeferencing of these features (X, Y) is highly accurate and the mine water head readings have accuracies of 0.01 m. The spatial distribution of the monitoring stations is uneven: for example, there are many monitoring points in Lothian and Fife but few in the Central or Ayrshire coalfields. Mine water levels can change significantly over time; they can vary diurnally (tidal response), seasonally, in response to pumping within the mine system or (especially) during post-closure mine flooding. The TCA mine water head data used in this study to create Dataset 4 were derived from autumn 2021 (i.e. current data at the time of this study).

The mine water discharge locations (n = 81) are X and Y coordinates where water drains from coal mine workings under gravity. Locations were recorded using a GPS device during fieldwork in November 2021 and published in Walls et al. (2022). The accuracy of the X and Y values is related to GPS error, which is usually ±3 m. The dataset is not exhaustive; there remain other unmonitored discharges in the MVS that were not captured as part of this research. Dataset 5 was used to provide another indicator of mine water head, which could be combined with Dataset 4. For Dataset 5, it is assumed that, at the location of the discharges, the mine water head is effectively at the elevation of the ground surface. The elevation (Z) value for each of the discharge locations was extracted from the digital terrain model (DTM, Dataset 6) for each of the mine water discharge locations in Dataset 5.

The digital terrain model (DTM) was compiled from 370 tiles of Ordnance Survey Terrain 5 at a scale of 1:10 000, July 2021 version. The raster tiles are 5 × 5 km with 5 m spatial resolution. The elevation data have a root mean square error (RMSE) of 1.5 m for urban areas, and 2.5 m for rural, moorland and mountainous areas (Ordnance Survey 2017).

For this study, four screening criteria were employed to identify the most promising locations for abstraction–injection well doublet exploitation of mine water geothermal systems. In theory, other screening data could be employed in future improvements of the tool if reliable data were available: these will be discussed later. The criteria have been implemented by manipulating the layers represented by datasets 1–6 in a GIS (ArcGIS by Environmental Systems Research Institute, Inc., Redlands, CA, USA) environment (Fig. 2).

For the development of an abstraction–injection well doublet, placing both wells into the same worked seam risks very rapid breakthrough of reinjected water (cool water, if the scheme is used for heating purposes) in the abstraction well, unless there is very significant horizontal distance between the wells (Loredo et al. 2017a, 2018) or some form of thermally attenuating barrier between the wells (e.g. goaf). It is regarded as beneficial if the abstraction and reinjection wells can be developed in different worked seams, thus achieving long, tortuous flow pathways by stratigraphic, rather than lateral, separation. It is a philosophy that has been employed at mine water geothermal schemes in Gateshead (Banks et al. 2022; Adams et al. 2023; Triple Point Heat Networks 2023). Some city-wide mine water district heating and cooling networks are not constrained to a single plot of land, and have thus been able to achieve both stratigraphic and lateral separation (e.g. Heerlen, The Netherlands: Verhoeven et al. 2014). In the case of many small–medium projects, however, the developer may be constrained to a relatively limited plot of ground.

Dataset 1 was analysed to determine the number of overlapping polygons across the study area using the processes outlined in the flow diagram in Figure 3. By this means, all areas where there were more than one worked seams (mostly coal but some oil shale, limestone and ironstone workings are included in Dataset 1) below the surface were identified (Fig. 4). Areas with none or a single worked seam only were rejected as unpromising for MWG.

Areas underlain by shallow mined workings may be subject to a subsidence or ground instability risk (TCA 2017). Drilling, pumping and reinjection operations could conceivably enhance this risk. For example, rising mine water head levels can result in millimetre-scale uplift across coal workings that may induce deformation at pillar edges. This process, along with thermal oscillations, may induce localized collapse (Todd et al. 2019). While the actual zone of collapse and compression above a longwall seam is variable and is a function of the width of the panel (Younger and Adams 1999), a commonly used ‘rule of thumb’ states that for every metre of coal abstracted, 10 m of overlying rock is potentially affected by subsidence (Healy and Head 1984; Bell 1986). Given that the UK hosts some coal seams greater than 2 m thick, with roadways possibly around 3 m, a 30 m value is suggested by TCA as ‘at risk shallow workings’ (Abbate 2016). The presence of shallow voids may also give rise to issues of loss of drilling flush when drilling through them to access deeper horizons. Moreover, shallow workings may already have been grouted for stability or require future grouting prior to development, and hence are unsuitable as a geothermal resource as void spaces become filled.

Once overlapping seams were identified, Dataset 2 was used to exclude areas from MiRAS where there are shallow (<30 m depth) workings below the surface (Fig. 3). The resulting polygons were rasterized in preparation for combination with rasters from other criteria. Figure 4 shows locations associated with Criteria 1 and 2 – multiple overlapping worked seams, where shallow workings are not present.

Mine water head is not inherently linked to the depth of worked coal seams: for example, at the UK Geo-Energy Observatories (UKGEOS) site in Glasgow, a static mine water head of 0.5–3 m below ground level (BGL) is recorded in mine workings at depths of c. 45–85 m BGL (Palumbo-Roe et al. 2021). It is, of course, possible that shallow worked seams are dry in areas with a deep mine water level and it may even be the case locally that perched water tables exist in shallow, poorly connected mine systems.

However, depth to mine water head is a crucial factor when considering the environmental risk, engineering and cost effectiveness of a MWG system. First, the energy expended for pumping (and thus monetary pumping cost) is directly proportional to the pumping head depth. In their analysis, Athresh et al. (2015) found that for 50 l s⁻¹, a pumping head of 10 m BGL incurs annual pumping costs of £3700, compared to £37 000 for a head at 100 m BGL. In terms of energy, assuming a pump efficiency of 55%, the first scenario would expend around 9 kW power, the second would expend 89 kW. The latter figure would probably represent an unacceptable pumping power expenditure to recover a thermal resource of perhaps 1 MW_th from a 50 l s⁻¹ discharge.

Thus, deep mine water heads are disadvantageous from an energy and cost perspective. Large pumping heads will also require larger and heavier pumps, which will require greater engineering costs and possibly greater borehole diameters. Very shallow groundwater heads can also be disadvantageous; reinjection will cause heads to rise in the injection borehole (Banks et al. 2022), possibly requiring pressured well heads and management of groundwater flooding risk around the wellhead. Reinjection operations with shallow groundwater heads also bear the risk of unexpected mine water emergence from old shafts and adits.

Points from Dataset 4 (mine water levels from TCA conservation shafts and boreholes) were combined with locations of mine water breakout at the surface (Dataset 5) to create a dataset of geolocated (X, Y) points, each associated with a mine water head (Z). This was then used to create a mine water surface elevation layer (Fig. 5). Empirical Bayesian kriging (Matheron 1960; Chung et al. 2019) was used to interpolate the mine water head between real data points as the weighted average of surrounding points. The kriging equation determines a weighting factor for each of the influencing points to minimize variance. It produces a surface that is the best linear interpolation for the available data. As a result, it lacks local or small-scale heterogeneities that could be present in the real potentiometric surface. The accuracy is more reliable in locations where there is a greater density of control points (e.g. East Lothian and Midlothian), and less so where points are few and distal. The resulting raster has been clipped to the extent of Carboniferous mined strata in Scotland, and the vertical difference between it and the surface level (DTM) was calculated. This formed a ‘depth to mine water head (m BGL)’ raster layer (Fig. 5), with calculated depths to mine water head mapped in 10 m increments; shallower values (0– 20 m BGL) are shown in shades of pink and deeper values (20–60 m BGL) are shown in shades of blue. The raster was clipped to exclude depths greater than 60 m BGL. This depth cutoff is admittedly somewhat arbitrary but a static water level of 60 m BGL could easily lead to pumping water levels of 80–100 m BGL and thus large parasitic power pumping losses. Thus, the blue shaded zone represents a zone of (in the authors’ opinion) optimal mine water heads for both pumping and reinjection. The pink zone represents shallow mine water heads, where difficulties with reinjection may be experienced (but where an operator may want to consider treatment and discharge to a surface water: Walls et al. 2021). Figure 5 also shows areas where the mine water head is predicted to be above ground level – this interpretation may be ‘real’ and represent an area characterized by mine water discharge. Due to interpolation errors, it is arguably more likely that these simply represent ‘very shallow mine water’.

Drilling deep geothermal boreholes with large diameters for a suitable pump is very costly in the UK; especially if materials are required (e.g. stainless steel casing) to resist the corrosive (saline, reducing, warm and H₂S-rich) environment prevalent in deep mine workings (Banks et al. 2022). Predictions of drilling costs performed by TownRock Energy (not available in the public domain) have indicated that drilling cost per metre increases significantly beyond 250 m, largely due to the higher mobilization cost and day rate for a suitably large drill rig (Diamond 2022 pers. comm.). A 250 m depth ‘cutoff’ is admittedly somewhat arbitrary and does not imply that projects deeper than 250 m are unfeasible (e.g. Heerlen, The Netherlands: Verhoeven et al. 2014). This cutoff was selected to meet the original intention of this study: to identify the economically and technically optimal areas for MWG. One could, of course, argue that increased mine water temperature with depth could justify deeper drilling, although preliminary analysis (Banks 2023) suggests that this may not be the case in an economy where drilling is costly and the value of heat is relatively low. For an MWG system coupled to a heat pump, the controlling factors for heat delivery are temperature change at the heat exchanger (⁠$Δ$T) (Banks 2012) and flow rate (Bailey et al. 2016). A higher mine water temperature would allow a greater $Δ$T or a modestly improved heat pump coefficient of performance.

‘In-seam level’ point values (Dataset 3) were converted from metres OD to metres BGL by subtraction from surface level estimates derived from the DTM (Dataset 6). The resulting ‘depth to worked seam’ (m BGL) points were assigned different symbols depending on whether they were ≤250 m BGL or >250 m BGL (Fig. 6). This was converted to a raster surface by kriging. Inverse distance weighting (IDW) was used, whereby values are calculated using a weighted average of the nearest points. The weights are proportional to the inverse of the distance between the data point and the prediction location raised to the power of 2. As a result, as the distance increases, the weights decrease rapidly and thus only the 12 nearest points were considered for each IDW. In areas where multiple seams are worked (Criterion 1), there can be depth differences of tens of metres or more between the various worked seams. If, within a given area, a grouping of mine workings has elevation points above and below the 250 m BGL cutoff, the weighted average produced by the IDW raster layer dictates whether the area is deemed above or below the cutoff.

The MiRAS aims to identify optimal MWG areas based on expected overall project cost and risk. Criteria 1 and 2 aim to mitigate project risk; Criteria 3 minimizes operational expenditure (OPEX), while Criterion 4 minimizes capital expenditure (CAPEX).

The raster layers from criteria 1–4 (C1–4) were combined to form a final raster layer. The ‘raster calculator’ tool was used to find areas that met all of the study optimization criteria. The algorithm used was as follows:

• IF location has multiple overlapping worked seams AND has no shallow workings <30 m BGL, THEN assign value C[1,2] = 1, ELSE assign value C[1,2] = 0.

• IF depth of mine water head >60 m, THEN assign value C3 = 0, ELSE assign value C3 = ‘depth to mine water’ (in 10 m increments to 60 m BGL).

• IF depth to workings ≤250 (m BGL), THEN assign value C4 = 1, ELSE assign value C4 = 0

  $Raster output=[1,2](1 or 0)×C3(value of depth m BGL)(1)×C4(1 or 0)$

Equation (1) has thus generated a value representative of the predicted mine water head only where there are overlapping seams at an average depth of less than 250 m BGL, with no shallow workings present.

The surface MWG resources are symbolized corresponding to their origin (e.g. existing treatment schemes: active, passive; gravity fed or pumped) and the symbol size is related to the magnitude of G.

Table 2 and Figure 7 show the outputs of the MiRAS GIS-based methodology as areas of the SMV that are judged optimal for mine water geothermal exploitation, by means of abstraction–injection well pairs, based on the following criteria:

• overlapping seams, without shallow workings;

• mine water head between 0 and 60 m BGL; and

• average depth to workings of less than or equal to 250 m BGL.

Cumulatively, there is a total of 370.2 km² across 19 local authority areas that are judged optimal for MWG development (Table 2). North Lanarkshire comfortably hosts the largest optimal area of 90.9 km². The Supplementary material contains output for each local authority area, or the output can be viewed online at https://www.spatialdata.gov.scot/geonetwork/srv/eng/catalog.search#/metadata/63ccefed-0165-461d-a5a5-025b0b2463c5 or added to the Spatial Hub preview map: https://maps.spatialhub.scot/data_preview_map/.

In Figure 7, optimal areas for MWG are shown; the colour scheme corresponds to the predicted mine water head. The most densely populated area with MWG potential stretches between SE Glasgow, Wishaw and Airdrie (see Fig. 8), with mine water heads largely 0–20 m BGL. Other optimal sites, with mine water heads at 20–60 m BGL, are primarily located beneath densely populated centres including Ayr, Kilmarnock, Bathgate, Stirling, Alloa, Cowdenbeath and Lochgelly, together with smaller clusters of the towns in Midlothian, East Lothian and along the Fife coast to the NE of Kirkcaldy.

Figure 9 shows an example of the locations of surface MWG resources (gravity mine water drainage or mine water pumping stations) overlaid on the MiRAS raster maps. Their label number for the surface MWG resources corresponds to the reference number in the supplementary material of Walls et al. (2022). Specifically, Figure 9 shows the western extent of the East Lothian local authority area, hosting the Blindwells treatment scheme (#3 from Walls et al. 2022): a surface MWG resource with an estimated heat availability of G = 6.9 MW_th. It is regarded as the most promising mine water geothermal resource in Scotland (Younger 2012; Bailey et al. 2016; Walls et al. 2022), being in proximity to urban areas and extensive local development plans (Optimised Environments Ltd 2020).

This type of map allows the identification of local authorities with the greatest potential for MWG (Table 3). North Lanarkshire comfortably hosts the largest area of subsurface MWG resource, suitable for abstraction–injection well doublets, at 90.9 km². The local authority areas of Fife, East Lothian and West Lothian have the highest heat availability via existing surface gravity or pumped discharges, largely due to them hosting three of the largest Coal Authority pumping and treatment schemes (Frances, Blindwells and Polkemmet: Chen et al. 1999; Nuttall and Younger 2004; Younger 2012; Wyatt et al. 2014; Zebec Biogas 2022), which represent a combined thermal resource of 14.4 MW_th.

The MiRAS output for locations of three recent mine water geothermal systems or research sites has been compared with empirical findings from those sites to provide some quality assurance that MiRAS is, indeed, producing realistic output in mine void depth, mine water depth and geothermal potential.

The Shettleston Housing Association (SHA) mine water geothermal system operated for c. 20 years from 1999 in eastern Glasgow (55.8504° N, 4.1668° W) (Banks et al. 2009; Walls et al. 2021). It was a relatively small abstraction–injection doublet scheme operating between two different stratigraphic horizons. Although the Shettleston area has portions that meet all four MiRAS criteria, the exact location of the abstraction and reinjection boreholes does not meet all the criteria (Figs 10 and 11); specifically, it fails Criterion 1 (presence of overlapping seams). This conforms with the findings of Banks et al. (2009, 2019), who suggest that the abstraction borehole is likely to have been completed in the workings of the Ell Coal Seam of Westmuir Pit (dating from 1845–62) but speculate that the reinjection borehole may have been completed in unworked Coal Measures strata. It should be noted that the workings date from before the mandatory requirement to record mine workings (Younger and Adams 1999), and thus some coal workings may be erroneously mapped or unmapped. The mine water head at Shettleston is predicted by MiRAS to be 0–10 m BGL. This is very close to the head of 12–13 m BGL recorded in 2016 beneath SHA (Walls et al. 2020). This data point was not used to construct the MiRAS mine water head surface due to its early date.

The UK Geo-Energy Observatories (UKGEOS) Cuningar site is also located in eastern Glasgow (55.8383° N, 4.2008° W), and is a research facility for monitoring, testing and innovation of mine water geothermal energy systems (Monaghan et al. 2021). It is extremely well documented and is underlain by seven worked coal seams from the Farme Colliery dating between 1805 and 1928 (Findlay et al. 2020). There are five boreholes that intersect mine workings of the Glasgow Upper and Glasgow Main coal seams (Monaghan et al. 2019). The static mine water head (0.5–3 m BGL) from UKGEOS was used as a data point to construct the MiRAS mine water head interpolation. There are no shallow worked coal seams (<30 m) at UKGEOS and the highest seam is c. 45 m BGL. MiRAS correctly identifies the Cuningar site as an optimal location for an abstraction–reinjection MWG system, although the shallow mine water head would need to be carefully managed to avoid mine water breakout during reinjection (Fig. 12).

The Dollar site (56.165° N, 3.662° W) was researched by Walls et al. (2023) to evaluate whether ground stability investigations could be combined with data collection on mine water geothermal potential. The site has four seams worked during the last few centuries, with the most recent workings forming part of the Dollar Colliery in the 1950s and 1960s. The depths range from surface outcrop to 50 m BGL, extending to greater depths north of the site in a small, isolated syncline. A mine water discharge from the colliery was used as a controlling data point for the mine water head interpolation (Criterion 3); thus, MiRAS represents the mine water head in the Dollar Colliery with good accuracy. The terrain rises steeply north of the Dollar colliery, and the mine water head thus becomes deep within a short distance of the discharge. As much of the site is underlain by shallow workings (<30 m BGL) or single worked coal seams (Fig. 13), MiRAS only shows a small fraction of the site with overlapping workings that satisfies all four criteria for optimal MWG exploitation (Fig. 14).

The identification of optimal sites for mine water geothermal energy systems is a first step towards increasing uptake of the resource across Scotland. It is recognized that MiRAS is not a perfect interpretation of subsurface conditions but it does provide an excellent tool for stakeholders and decision-makers, allowing rapid screening of any site across the MVS for its mine water geothermal feasibility.

The feasibility of a MWG site is influenced by more than the four criteria generated in this study, although these are seen as the principal factors. Other identified controlling ‘geofactors’ that could conceivably be added to a subsequent version with a multi-criteria evaluation technique are:

1. Mine water temperature: Farr et al. (2021) have demonstrated that mine water temperature, at least under static conditions, is related to depth. Thus, the ‘depth of working’ information on which MiRAS is based could be used as a proxy to predict mine water temperature. Implementation of this would require careful consideration when dealing with multiple worked horizons as, given multiple worked seams, groundwater could be abstracted from a deeper seam and reinjected to a shallower (to maximize abstraction temperature) or vice versa (to minimize risk of uncontrolled mine water surface emergence). Temperature is thus a 3D variable that would be challenging to represent in 2D tool such as MiRAS. Moreover, it must be recognized that mine water temperature could potentially change once an MWG scheme becomes operational and mine water actively starts to circulate within workings.

2. Mine water hydrochemistry: Walls et al. (2022) have compiled a thorough overview of mine water chemistry from a number of surface discharges of mine water throughout the MVS. Mine water chemistry can be highly influential in the operational cost or long-term sustainability of an MWG scheme, High concentrations of dissolved iron can lead to the clogging of wells, pipes or heat exchangers with ferric oxyhydroxide, while highly reducing H₂S-rich saline waters can lead to corrosion risk, even of some stainless steels. Unfortunately, it is recognized that mine water hydrochemistry can be highly stratified, with both iron-rich, incrusting, mine water and saline, reducing, corrosive mine water existing at a single site, within different mined horizons (Banks et al. 2022). Such a complex 3D variable, such as hydrochemistry, thus cannot be readily represented in a 2D tool such as MiRAS, and the hydrochemical data collated by Walls et al. (2022) may not be a good guide to the hydrochemistry at depth.

3. Type of working: it is recognized that mine water geothermal exploitation requires different exploration strategies in different types of workings. For example, in old pillar and stall workings, the accuracy of old plans may not allow the reliable targeting of voids; rather, a statistical approach might be needed (e.g. 50% chance of success if the void to pillar ratio is around 1:1). In more modern longwall workings, it may be more appropriate to target either access roadways to longwall panels or fractured/collapsed zones above longwall workings (Andrews et al. 2020). TCA data do contain information on dates of working for seam sections. Dates of working, possibly combined with an interpretation of working geometry, might conceivably be used to estimate the likely method of working (pillar and stall, total extraction or longwall panels), although there was considerable overlap of working methods throughout time, and geometrical interpretation would be time-consuming.

4. The plotting of existing open mine shafts, which might provide access to workings (e.g. as one pole of a doublet), without the additional cost of drilling or which might be used as ‘standing column’ heat exchange systems, where water circulation takes place within a single shaft (as described by Burnside et al. 2016 at Markham, Derbyshire, UK). While TCA does possess knowledge of the state of backfill/openness of some shafts, the state of many others in not documented with certainty.

5. The annual rainfall and surface topographical gradient could be incorporated into MiRAS as proxies for hydraulic gradients and throughflow rates in mine workings.

6. Regular updating of mine water head surface: Dataset 4 (monitored mine water head) was a ‘snapshot’ of mine water head data held by TCA in autumn 2021. While post-abandonment mine water levels have largely recovered across the MVS, there are still some mines where recovery is in progress. There are other locations where head may be affected by variations in TCA pumping regimes. Ideally, Dataset 4 within MiRAS should be updated about every 5 years to reflect possible systematic temporal changes in head.

Moreover, TCA do possess other datasets that have not been incorporated into this version of MiRAS but could be included in future versions. For example, TCA hosts:

• A more extensive polygon layer that shows ‘Probable shallow workings’ – i.e. areas where coal was present at shallow depths and TCA believe that working is likely to have occurred prior to mandatory plans. This could feasibly be used to supplement Dataset 2 (known ‘Shallow workings’).

• An ‘In-seam contour’ dataset, derived from mine plans where seam elevations have been contoured. This could be used to supplement Dataset 3 (‘In-seam level’).

The results should be viewed in light of some data source limitations and wider MWG considerations pertaining to the development of these resources. Some of these limitations and uncertainties are as follows:

1. Data density: in some areas, a low density of data points in some locations leads to inherent uncertainty in interpolated data, especially in the case of the predicted mine water head.

2. Number of overlapping seams: MiRAS at present merely flags locations where multiple worked seams are present; it does not indicate the specific number of overlapping worked seams.

3. Areal size of mine water resource: MiRAS indicates areas where four specific MWG criteria are satisfied. In some cases, these areas may be quite small and may give the impression of a very limited geothermal resource. In some cases, however, such limited areas may be hydraulically linked to much more extensive areas of mine workings where, for example, only two or three of the criteria are satisfied but which may form part of a larger mine water resource. Clearly, MiRAS should not be understood as a ‘resource map’ per se but as a map indicating locations where a resource can be optimally accessed via drilling. For example, in the case of the Dollar Colliery site, only a small area meets all four MiRAS criteria (Fig. 14) but the extent of interconnected flooded workings forming the resource (Fig. 13) is considerably larger.

4. Exclusion of very shallow and very deep workings: the MiRAS concept might be criticized on the grounds that the ‘depth of workings’ cutoffs for shallow (<30 m deep) and very deep (>250 m) workings are somewhat arbitrary. Indeed, the relatively shallow workings of the Dollar Colliery (Figs 13 and 14) have been described in detail by Walls et al. (2023), who indicated that there might, indeed, be some potential for limited MWG development at the site, although that this should be in coordination with ground stability (and possible future ground stabilization via grouting) assessments. Others (Farr et al. 2021) have pointed out that the highest mine water temperatures typically occur in the deepest workings and that these might be regarded as the most attractive prospects. In this paper, we have argued that exploration for and exploitation of such deep resources comes with significantly elevated exploration and drilling costs, and also with potentially undesirable (saline and highly reducing) mine water chemistry. The authors do, however, recognize that the ‘cutoff’ at 250 m is arbitrary and might be revised in future iterations of MiRAS if drilling costs fall and if the understanding of resource exploitation in deep workings improves.

5. Focus on newly drilled, single well doublet systems: it is acknowledged that MiRAS assumes that the most likely MWG exploitation methods will be via newly constructed borehole doublets drilled into different worked seams at a single location or in close proximity. There are other methods of MWG exploitation that MiRAS does not explicitly consider. Amongst these are: (i) single borehole(s) for abstraction only, followed by heat exchange, water treatment and disposal to a surface watercourse; while technically feasible, it is assumed that these will be unattractive on grounds of ongoing treatment cost; (ii) abstraction–injection doublets but at widely spaced locations, as is implemented at Heerlen (The Netherlands); such systems are undoubtedly attractive, but the ownership of widely spaced land parcels and access to a pipeline corridor between them is likely to require development by a regional authority (such as a council or municipality); and (iii) standing column solutions, where water is circulated only within a single mine shaft or borehole, of the type described by Burnside et al. (2016) at Markham, near Bolsover, Derbyshire.

MiRAS provides a valuable first-pass screening tool for interested individuals prior to, or at an early stage of, project development. It cannot, however, replace the necessary expertise required for development of a MWG system. It contains little internal detail on mined areas and cannot be used to identify exact borehole locations, probable underground flow pathways or resource size. Successful project development requires integrating skills of hydrogeologists, mining engineers, chemists, heating, ventilation and air conditioning (HVAC) and buildings services engineers, economists, and planners (Walls et al. 2021). Successful design and development of MWG requires a good understanding of underground flow pathways to ensure a sustainable system and to mitigate the risk of thermal ‘feedback’ and depletion. The risk of thermal feedback is present in open-loop doublet MWG systems that are heavily dominated by either heating or cooling provision (MacAskill et al. 2015). Modelling techniques are available to identify and predict the rate and timing of thermal feedback (Ghoreishi Madiseh et al. 2012; Loredo et al. 2016, 2017a, 2018; van Hunen et al. 2022). However, there is an increasing recognition that development of mine water-coupled heating and cooling networks where there is a long-term balance between heat extraction and heat rejection, which use mines as a thermal buffer (Verhoeven et al. 2014; Fraser-Harris et al. 2022) or which use mines for underground thermal energy storage (UTES), would ultimately more likely to be sustainable. Specialist hydrochemical and water quality engineers are also required to design wells, pipe systems, heat exchangers and maintenance programmes that minimize the risks of clogging and corrosion. While district heating and cooling networks will require a certain density of demand to be viable (which can be problematic in rural areas: James Hutton Institute 2016; Harnmeijer et al. 2017), recent requirements for local authorities to consider ‘heat network zones’ can only encourage consideration of innovative regional environmental thermal resources (Scottish Government 2023b; DESNZ 2024).

Surface mine water (gravity or pumped) discharges, where no reinjection of ‘thermally spent’ water is undertaken, are at little risk of thermal feedback and can thus be exploited in strongly heating- or cooling-dominated modes to satisfy potentially large heating and cooling demands (TCA 2020; Wood and Crooks 2020). The water will usually require treatment prior to discharge to surface water (Banks et al. 2019), however, and consideration should be given to temperature changes (due to heat extraction or rejection) impacting the efficacy of treatment processes or the ecology of recipient water bodies (SEPA 2016, 2022).

Data on abandoned mines in the MVS, and the waters they contain, has been compiled from TCA and Walls et al. (2022). These data have been combined with surface elevation data from the Ordnance Survey to create a Mine Water Geothermal Resource Atlas for Scotland (MiRAS). The datasets were processed in a GIS environment to produce interim GIS layers that show locations where:

1. there exist overlapping worked coal seams;

2. there are no shallow (<30 m BGL) workings that might adversely affect ground stability;

3. mine water head is within 60 m of the surface (as predicted by kriging mine water elevation points); and

4. depth to mine workings is less than 250 m.

The four criteria were combined, and the resultant map indicates an area of 370.2 km² optimal for mine water geothermal development. These optimal areas can be found throughout the MVS, with the greatest footprint of potential sites in North Lanarkshire.

Ground-truthing of the MiRAS tool at three sites (Shettleston, Cuningar and Dollar), where mine water geothermal resources have been exploited or researched in great detail via intrusive ground investigation, indicates that MiRAS does, indeed, give a representative ‘first-pass’ indication of MWG potential, lending confidence in its utility as a screening tool.

The MiRAS allows areas to be screened for mine water geothermal (MWG) potential, speeding up project initiation and empowering potential ‘champions’ of mine water geothermal energy. Areas that are denoted for residential, industrial or commercial development in local plans can be cross-referenced with the MiRAS to understand whether or not MWG is a potential thermal supply. It is envisaged that this new means of rapid site appraisal will prove to be a useful tool for local authorities, landowners and other stakeholders when exploring low-carbon heating opportunities and, as a result, expand the awareness and increase the uptake of MWG resources. Future versions of the MiRAS could be expanded to include other factors, such as water chemistry or depth intervals, and have its impact improved by coupling to databases regarding current and future heating and cooling demands. MiRAS cannot, however, replace the need for subsequent detailed technical specialist involvement in the design and development of MWG.

The authors would like to thank The Coal Authority for the provision of datasets that were invaluable to this study. The authors also extent their thanks to the Improvement Service for hosting the resultant Mine Water Geothermal Resource Atlas for Scotland online. The authors would also like to thank the reviewers for comments which helped to improve the manuscript.

DBW: conceptualization (equal), data curation (lead), formal analysis (lead), investigation (lead), methodology (lead), project administration (lead), writing – original draft (lead), writing – review & editing (equal); DB: conceptualization (equal), methodology (supporting), supervision (equal), writing – review & editing (equal); YK: methodology (supporting), supervision (supporting), writing – review & editing (equal); AJB: supervision (equal), writing – review & editing (supporting); NMB: conceptualization (supporting), methodology (supporting), supervision (equal), writing – review & editing (equal).

N.M. Burnside was funded by a University of Strathclyde Chancellor's Fellowship, A.J. Boyce was funded by the NERC National Environmental Isotope Facility award at SUERC (NEIF-SUERC, NE/S011587/1) and part of the analysis was funded by the John Mather's Rising Star Award awarded to D.B. Walls.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

The datasets generated during and/or analysed during the current study are available on the Improvement Service's Spatial Hub platform (https://www.spatialdata.gov.scot/geonetwork/srv/eng/catalog.search#/metadata/63ccefed-0165-461d-a5a5-025b0b2463c5). Enlarged maps of each affected local authority area are included in the Supplementary material.