Development of a drought-resilient water supply from dolomitized limestones of the Irish Midlands

Groundwater is a critical source of drinking water globally, making up half of the drinking water consumed in Europe, for example (Völker and Borchardt 2019). Groundwater stored in aquifers is also an important buffer in the water cycle, attenuating and delaying droughts, and thereby reducing drought-induced loss (Cuthbert et al. 2019). Groundwater relies on renewal through recharge (i.e. complex processes that lead to the refilling of the groundwater source from precipitation: Van Lanen and Peters 2000). Such recharge processes are impacted by periods of drought, yet the available literature linking drought and groundwater is not extensive (Petersen-Perlman et al. 2022). In temperate climates, where over one-third of the world's population live (Klinger and Ryan 2022), recent research points to drought vulnerability (e.g. Holman et al. 2021) despite such regions historically not being considered at major risk of drought impact.

Water supply in Ireland reflects this global context, with additional global issues such as population growth, ageing infrastructure and climate change acting as further stressors (Irish Water 2021). This study concerns the development of a water supply for a town in the Midlands of Ireland. The study objectives included: (i) identification of the water source to the groundwater supply wells; and (ii) investigation and explanation of the resilience of the source during drought. The resilience of new and operating water supplies is critical in the global context described above, and this study contributes to knowledge on how resilience can be assessed for groundwater sources. This study approaches water supply at the subcatchment scale (as also used by the European Water Framework Directive (WFD)), a scale considered optimal in water resource planning (Nørskov Gejl et al. 2020) but under-represented in the literature compared to other scales (Barthel et al. 2021).

In 2007, the local authority (LA) for County Laois, in the Irish Midlands, identified a need for a new source of at least 1500 m³/day, and up to 4000 m³/day, to cater for regional growth at Graiguecullen (part of the Carlow town urban area, with a 2016 population of about 25 000). Due to their more stable water quality, water supplies developed from high-quality groundwater sources generally demand less treatment than those from surface waters (EPA 2020), and in this case were preferred by the LA. The LA also preferred for any new groundwater source to have low levels of nitrate, and be within an economical distance of the customer base.

To identify a potential water source of the required size, the LA first commissioned a hydrogeological investigation to improve the understanding of conditions west of the River Barrow at Carlow, details of which are presented here. Once a source was identified, the LA developed it into an operational water supply in 2012. This study assesses how the groundwater levels at the production wellfield responded to drought conditions throughout a period of drought in 2013. Irish Water now manages this water supply as part of a single water resource zone that covers much of the Carlow area, and considers resource planning for the area cumulatively.

Carlow town lies on the River Barrow, Ireland's second-longest river, which drains southwards into Waterford estuary. Upstream of Carlow, the River Barrow drains a catchment of about 2000 km² of mainly arable land. At Carlow (60 m above sea level), the Barrow Valley floor is about 8 km wide, bordered to the east by hills of Leinster Granite and to the west by the uplands of the Castlecomer Plateau.

The study area within this overall setting was constrained by LA administrative boundaries, suitable geology and distance to the demand centre. This results in a study area of about 11 km² to the NW of the town (see Fig. 1) but the catchment area extends beyond this, so that water flowing within the study area originates in part beyond the LA boundary. The study area topography is flat and undulating.

The study area is part of a relatively dry and warm region within the Irish cool temperate climatic zone (Tedd et al. 2012). Climate data are recorded by Met Éireann at Oak Park synoptic station, about 3 km east of the study area (Fig. 1). Annual long-term average (LTA) rainfall at the station is 840 mm and annual potential evapotranspiration (PE) is typically about 550 mm. A study of historical drought in Ireland (O'Connor et al. 2022) confirmed the importance of groundwater storage and catchment characteristics in mediating drought incidence and propagation of meteorological to hydrological drought. That study characterized Irish catchments into three clusters, based on drought response. The Barrow Valley catchment belongs to the cluster that exhibits the fewest hydrological droughts but, when they do occur, they tend to last longer and produce greater accumulated and reduced mean deficits than other cluster types (O'Connor et al. 2022).

The study area is drained by the Sleaty stream, which flow eastwards to the River Barrow, and the Fushoge River, which flows southwards (Fig. 1). These watercourses occupy parts of two surface-water subcatchments (on the western side of the River Barrow). The subcatchments are defined by Environmental Protection Agency (EPA) implementation of the WFD as Barrow_sc_070 (incorporating the Douglas River and Sleaty stream) and Barrow_sc_110 (incorporating the Fushoge River catchment) (EPA 2021).

Relatively intensive agricultural land use (including a relatively high proportion of dairy and tillage) across much of the centre and east of the River Barrow catchment is linked to a history of high nitrate levels in surface water and groundwater (Daly 1981; EPA 2021). West of the River Barrow, relatively lower-intensity agriculture predominates (tillage and cattle grazing), and significant river agricultural pressures are not identified in the study area. The Sleaty stream is deemed to be moderately impacted by hydromorphology pressures but agricultural pollutant pressures on the river were not identified at this time. The Fushoge River in the study area is deemed to be of good status (EPA 2021) but downstream reaches of the Fushoge River (Barrow_sc_110) are deemed to be affected by significant river agricultural pressures.

The general stratigraphy of the area (Table 1) comprises a Dinantian limestone sequence that is unconformably overlain by Namurian and Westphalian sandstones and shales (Tedd et al. 2012). The Namurian and Westphalian sandstones and shales form the Castlecomer Plateau. The Barrow Valley is floored by a sequence of Dinantian pure bedded limestones and dolomitized limestones (shown in bold in Table 1) that extend north and south (Fig. 1). The 1:100 000 scale geological map (Tietzsch-Tyler and Sleeman 1994) for the region classifies the Dinantian limestones underlying the study area to be of the Ballyadams Formation, with the Clogrenan Formation along the western margin of the study area. The Ballyadams Formation is a thick-bedded crinoidal limestone where bedding is often separated by ‘clay wayboards’ resting on eroded and karstified surfaces. The Clogrenan Formation is distinguishable from it by the absence of clay wayboards and the presence of chert. The dolomitized limestones of the Milford Formation lie east of the study area.

Bedrock is overlain by varying thicknesses of superficial deposits from the Quaternary Period. Regionally, fluvioglacial outwash from glacial retreat from the last glaciation produced the large sand and gravel body that occupies much of the Barrow Valley west of the study area (Fig. 1). Within the study area, the underlying subsoil is predominantly a till derived from Carboniferous limestone but minor alluvial deposits and sand and gravel bodies are also present. Relatively deep drains are excavated along farmland in the area, indicating lower-permeability soils derived from the glacial till (Conry 1987).

A large zone of ‘regional dolomite’ (dolomitized limestones) in the Tournaisian limestones (Table 1) of the Irish Midlands is thought to extend, discontinuously, through several parts of central Ireland (Mulhall and Sevastopulo 2004), including the study area. In the Carlow area, thick glacial superficial deposits obscure the exact extent of the regional dolomite. Regional maps indicate that dolomitized limestones occur mainly on the eastern side of the valley, and include the lower half of the Milford Formation and the Ballysteen Formation (Tietzsch-Tyler and Sleeman 1994).

About 5 km east of the study area, the Leinster Granite, a large batholith comprising several smaller granite intrusions (Daly 1981), crops out (Fig. 1). The Dinantian limestone strata dip gently (5°–15°) to the west, off the Leinster Granite and under the Castlecomer Plateau (a large syncline). Thick glacial superficial deposits in the valley mask much of the evidence of faulting in the Carboniferous strata but faulting is assumed to be present to a significantly greater degree than shown on published mapping.

At Hollymount, about 1.5 km NE of the study area (identified in Fig. 1), there is a large, infilled karst feature in the limestone (Parkes et al. 2016). A Geological Survey Ireland (GSI) research borehole there recorded a thickness of Neogene-aged sediment – about 30 m of ‘blue grey clay’ that becomes black with depth – beneath about 18 m of Quaternary glacial till (Parkes et al. 2016). The borehole was stopped in weathered limestone shale at about 48 m. Palynological analysis of the organic sediments indicate a Miocene or earliest Pliocene age (Coxon and McCarron 2009). The Hollymount site is considered significant since Paleogene–Neogene-aged sedimentary deposits are rare in southern Ireland, with few other sites documented (Coxon and McCarron 2009). The extent of the infilled karst feature is not known.

The two bedrock units of interest in the succession of Dinantian strata at the Barrow Valley are categorized together as a Regionally Important Aquifer (diffuse flow) under the GSI aquifer classification system (DELG–EPA–GSI 1999). These units are the dolomitized limestones and the karstified pure bedded limestones (Table 1). The dolomitized limestones are of particular interest to groundwater resource development because they have enhanced secondary porosity (Mulhall and Sevastopulo 2004; Nagy et al. 2004), and fissuring in dolomitized limestones can extend to more than 200 m in depth (GSI 2004). The known extent of dolomitization is obscured by the cover of superficial deposits in the study area but the proximity of the dolomitized limestones to the study area raises the possibility that dolomitization may extend into it.

The GSI public database of well records indicates that the highest yielding boreholes are within 500 m of the River Barrow, an area that intersects both the sand and gravel aquifer and the limestone aquifer. Well records do not elaborate on the source aquifer. Major abstractions are not recorded in the database owing to the relatively low-intensity agriculture on the western side of the River Barrow.

The relatively intensive agricultural land use across much of the centre and east of the River Barrow catchment is linked to a history of high levels of nitrate in groundwater (Daly 1981; EPA 2021). As a result, the main gravel groundwater body in the area is now considered ‘at risk’ under the WFD classification (EPA 2021). West of the River Barrow, and the gravel groundwater body, in the study area, relatively lower-intensity agriculture predominates (tillage and pasture).

National recharge maps indicate that the Barrow Valley is an area of relatively high (>500 mm a⁻¹) recharge (Hunter Williams et al. 2011). The presence of a regionally important sand and gravel aquifer overlying the limestones of the aquifer may intercept much of the recharge at a shallow level. The degree of interconnection between the sand and gravel aquifer and the underlying limestone units is not known but they are assumed to be interconnected in places. The presence of glacial till on the western side of the Barrow Valley may isolate the limestones from groundwater flow in much of the overlying superficial deposits and, although the limestone units are saturated, groundwater circulation in them may be low (GSI 2004).

Point recharge into the limestones on the western part of the Barrow Valley, via swallow holes at the base of the Castlecomer Plateau, is reported to occur as streams cross from Namurian-aged shale into Dinantian-aged limestones (GSI 2004).

In the study area, the combination of thick overlying glacial deposits and Carboniferous limestones with essentially no primary permeability poses a challenge to identifying suitable locations for drilling. In such cases, a staged strategy of reconnaissance, surface geophysics and pilot well drilling is suitable to try to identify zones of high (secondary) permeability (Misstear et al. 2006). The overall methodology used for the study for identifying the resource and developing it into a water supply followed four stages (Table 2).

The key field activities during Stage 1 were interviews with landowners in the area and understanding the groundwater conditions (borehole depths and yields, and groundwater level and quality) on their land. In the absence of significant industrial water users, the largest abstractions of groundwater were surmised to be from larger farms, so the well survey focused on these. Eight wells were identified for groundwater sampling and level monitoring (Fig. 1), and samples from these were analysed for field and common ion parameters. Samples were obtained from taps closest to these boreholes, once field parameters (temperature, pH, specific electrical conductivity and dissolved oxygen measured using a Hanna Instruments^® multiparameter probe) had stabilized.

The goal of the reconnaissance stage was to further delimit a suitable area for groundwater source development. Three sites (sites A, B and C) were shortlisted for more detailed investigation, based on the Stage 1 outcome.

Surface geophysics, using electromagnetic very-low-frequency resistivity (EM-VLF-R) and 2D resistivity profiles, was used to estimate depth to bedrock and image possible subsurface groundwater flow paths in the bedrock. The surface geophysics was carried out by Minerex Geophysics Ltd.

A total of 191 electromagnetic EM-VLF-R measurements were taken on a 100 m grid (where practicable) over an area of 144 ha at the three shortlisted sites to estimate the superficial deposits thickness and bedrock resistivity. A Geonics EM16R instrument with a resistivity attachment and two stainless steel probes with a 10 m cable were used to determine the apparent resistivity and the phase angle. The survey readings were all taken using a propagated wave from the British VLF station in Anthorn, Cumbria, UK (19.6 kHz). The depth of penetration is dependent on the recorded apparent resistivity (measured in ohm-metres (⁠$Ω$ m)) and therefore varies spatially with variations in the apparent resistivity.

The 2D resistivity survey specifications comprised a multi-electrode switching system: a TIGRE resistivity meter, laptop, power supply, a maximum of 128 electrodes under computer control per array, 5 m electrode spacing, maximum profile length of 635 m, Imager 5 cables, stainless steel electrodes, contact resistances of less than 2000 W m, a Wenner electrode configuration and three cycles per reading to reduce background noise. 2D resistivity profiles determined the subsurface resistivity on a cross-section, and profiles 635 m in length gave a depth penetration of 100 m. 2D resistivity data were inverted using the RES2DINV inversion package – this program uses a smoothness-constrained least-squares inversion method to produce a 2D model of the subsurface resistivities from the recorded apparent resistivity values. The Jacobian matrix was recalculated for the first two iterations, and a quasi-Newton approximation was then used for subsequent iterations. Each dataset was inverted using five iterations, resulting in a typical root-mean-square (RMS) error of less than 2.7%.

The first drilling phase took place in 2008, using rotary drilling with air flush, with locations guided by the surface geophysics results. The rotary drilling process does not allow the recovery of undisturbed samples, so all descriptions provided were interpreted from cuttings recovered during drilling. Apart from one pilot well (Site B), all wells were completed with a screen/casing string over their full drilled depth. Sanitary seals were installed using pressure grout methods. Observation wells were installed near pumping wells at two of the three sites. Well materials used during the 2008 drilling were PVC casing and slotted screen. Well ‘development’ activities, whereby surging and flushing with compressed air was performed while the rods were cycled up and down the well, were then carried out, which required a total of 23 h to achieve satisfactory clearing of suspended solids from the well. Following this first phase of drilling and well installation, pumping tests were performed. These involved the installation of a submersible pump to a nominal depth of 50 m in each pilot/production well and performing a standard programme of tests: calibration, step-tests, constant rate tests and recovery tests. Drawdown was measured using electronic pressure transducers and a well dipmeter (for redundancy), and discharge was measured using a digital flow meter. At the end of the constant-rate pumping tests, a sampling port from the discharge line was directed to a flow-through cell, where the water quality field parameters were measured using a Hanna Instruments^® multiparameter probe for temperature, pH, specific electrical conductivity and dissolved oxygen. Samples were taken at the sampling port, stored in a cooler box and dispatched the same day to a certified laboratory (ALcontrol Laboratories, Dublin; now called ALS). All other parameters were analysed at the laboratory according to standard methods.

In 2009, a further phase of drilling was carried out at Site A to install an additional production well and an observation well. During that phase, reverse mud circulation drilling was used for improved hole stability. The second production well was installed with a continuous wire-wound screen (Johnson Hi-Flow™ 304 well screen) for greater well efficiency. A similar exercise in well development was again performed at the new well until satisfactory clearance was achieved. During 2010, a single-well test at the new production well and a combined pumping test involving both of the Site A production wells were carried out. Information from the pumping tests was used to estimate recharge boundaries using image well analysis (Domenico and Schwarz 1997). Field water quality measurements and water samples were again taken, following the same protocol as outlined above.

Throughout the feasibility assessment period, groundwater levels were measured periodically using a groundwater level meter. The time series of these was assessed alongside effective rainfall (the rainfall potentially available for groundwater recharge). As with regional studies of recharge (Tedd et al. 2012), effective rainfall was calculated as the difference between the total rainfall and the actual evapotranspiration via the Penman–Monteith soil moisture budgeting method. Climate data were provided by Met Éireann for its Oak Park synoptic station (licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) License).

At Site A, construction of the required water supply infrastructure (reservoir and pumping station) took place during 2011, and the facility began delivering water to the municipal network in 2012. No groundwater data were collected over that period.

A system control and data acquisition (SCADA) system began data collection in January 2012 at the water supply facility, capturing commissioning and operational phases. Data are recorded every 10 min, including the hydrostatic level and flow rates from both production wells.

The pumping regime generally involved one reservoir filling cycle per day: pumps operated simultaneously and, on average, only needed to operate for about 60% of the time to complete the filling cycle. A daily maximum and minimum were extracted from the SCADA hydrostatic level dataset to represent non-pumping groundwater levels and pumping/dynamic groundwater levels, respectively. The hydrostatic level was reduced to a nominal reference level, set at 2 m during the January 2012 commissioning period. Time-series plots of the operational period for groundwater level and effective rainfall were generated for comparison with Stage 2 data. Groundwater sampling and analysis data for water quality were provided by the LA laboratory.

The resilience of the pumping groundwater level to periods of drought was assessed using a graphical method (Misstear and Beeson 2000). This method uses the operational data collected from the SCADA system. Aside from idealized analytical methods (using step-test information) that are often used for determining sustainable yields, other methods have been developed for predicting groundwater levels during drought, typically using meteorological variables-based drought indices (e.g. Leelaruban et al. 2017). Such methods achieve reasonably good correlation in many cases and are suitable for rapid assessment. However, in the case of this study, the graphical method (Misstear and Beeson 2000) was preferred because: (1) it uses easily collected and, in this case, automated metrics for a water supply (i.e. groundwater level and flow rates); and (2) in dealing with a heterogeneous aquifer, it is better to rely on operational data rather than idealized analytical approaches that assume uniform conditions over thick homogenous aquifers.

Interviews with agricultural landowners revealed a maximum depth of drilling of 30 m for the eight wells inventoried (Fig. 1) but little information on the ground conditions. Five of the eight wells were dug wells (probably no deeper than 8 m), with the remaining three being bored wells (Table 3).

The shallow well depths mean that water quality results from private wells are unlikely to reflect deep aquifer groundwater quality. The pattern of variable nitrate concentrations and the incidence of high nitrate levels reflect the regional conditions in the Barrow Valley (e.g. Daly 1981; EPA 2023) and the SE region of Ireland, which has the highest concentrations of the five regions assessed nationally (EPA 2023). Nitrate levels at the study site were below 10 mg l⁻¹ NO₃ at wells 3, 5 and 6, in areas mapped as glacial till rather than sand and gravel (Fig. 1). It is thought that groundwater beneath well-drained soils (derived from sands and gravels) may be more prone to receiving dissolved applied nitrate fertilizer during recharge events, as noted during another nearby study (Premrov 2011).

Despite being a useful means of stakeholder engagement (Mott Lacroix and Megdal 2016), the initial reconnaissance of local private water use was of limited effectiveness here in identifying large water resources – both because there were few wells in the sparsely populated area, and most supply requirements are of small scale and are therefore satisfied by accessing only the shallow (superficial deposits) aquifer.

Three land parcels were therefore shortlisted based both on the willingness of stakeholders to participate and low nitrate levels in the overlying aquifer. A follow-on surface geophysical survey was designed to target the vicinity of low-nitrate water quality and potential discharge zones tentatively identified. The extent of these land parcels are shown as sites A, B and C in Figure 2a.

Broad interpretation of the grid of EM-VLF-R surface geophysics was in three zones (Fig. 2a). These zones are: (1) high phase angle (>45°)/low apparent resistivity, indicating a thick overburden or highly karstified or dolomitized limestone; (2) low phase angle (<45°)/high apparent resistivity, indicating relatively shallow and competent (i.e. less karstified) limestone; and (3) an intermediate transition zone. VLF-EM-R findings were supported by information from the 2D resistivity sections: the interpretation of three of the main 2D resistivity lines (Fig. 2b) was confirmed with pilot well-log geological information (Fig. 3).

For the wellfield at Site A, wells are prefixed according to well type (P for production well and O for observation well: Fig. 4). The logs of two observation wells at Site A (within 30 m of each production well) did not differ markedly from the production well logs, so are not shown. For sites B and C, pilot wells are identified by their site code only.

In contrast to the reconnaissance stage, the combination of geophysical methods (VLF-EM-R and 2D resistivity) proved effective in subsurface characterization. The efficacy of 2D resistivity for this application in Ireland has been previously documented (e.g. McCormack et al. 2017) but VLF-EM-R was also found to be effective in indicating the surface extent of the large-scale karst structures in a cost-effective manner (once this was correlated with the 2D resistivity and pilot well data).

For line R-A in Site A (Fig. 2b), a probable normal fault structure dipping at 60° was interpreted to be present, extending deeper than the depth of the 2D resistivity pseudosection (100 m). The apparent resistivity map for Site A (Fig. 2a) also indicates a lower-resistivity feature traversing the site along a NNE trend. Unaltered bedrock within the area is assumed to form part of the Ballyadams Limestone Formation but was completely altered to dolomite at PA-1 (Fig. 3, drilled over the interpreted fault structure). Dolomitization appears to be associated with the fault zone – this relationship is seen elsewhere in the ‘regional dolomite’ areas of the Irish Midlands (Dodds et al. 1994). The presence of dolomitized limestone here adds to instances of regional dolomite in the Irish Midlands (Mulhall and Sevastopulo 2004), and possibly connects to the regional dolomite mapped near the granite (Nagy et al. 2005) in the eastern part of the Barrow Valley (Fig. 1).

Depth to bedrock (21–27 m depth where logged at wells) suggested thicker glacial infill along the relatively weak altered bedrock of the fault zone. Depth to bedrock appeared to shallow considerably over short distances (apparent on 2D resistivity sections: Fig. 2b) to less than 5 m in the eastern part of Site A. This depth to bedrock variability is characteristic of many Irish geological settings and adds to the complexity of identifying a groundwater source that balance low vulnerability with receiving sufficient recharge (Comte et al. 2012). The overlying superficial deposit was predominantly glacial till, which confined the groundwater in the dolomitized limestone – major groundwater ingress was at the top of bedrock, with the static level rising to within 2–3 m of ground level.

A substantial low-resistivity feature more than 200 m in width is indicated on the 2D resistivity sections R-B and R-C (Fig. 2b, as well as other 2D resistivity sections not detailed here, and EM-VLF-R results), extending beyond the maximum 2D resistivity penetration depth (100 m). The feature is termed here the Clonmore feature, after the local townland. Depth to bedrock from geological logs supports evidence of an infill feature – depth to bedrock at sites B and C was considerably deeper (42–61 m depth where logged at wells) than at Site A. The geological log of well B-1 (Site B) comprises a stiff grey clay, transitioning to a black carbonaceous stiff clay down to 48 m. Bedrock (mainly chert), interpreted to be highly karstified limestone, was recorded below a depth of 48 m. Collapsing hole conditions during drilling at Site B meant that the borehole could not progress more than a few metres into underlying strata. Similarly, in parts of Site C, thicknesses of more than 50 m of stiff dark grey clay overlie interpreted bedrock. The geological descriptions and 2D resistivity results together indicate that the Clonmore feature is an infilled karst hollow that is likely to extend north–south for at least 3.5 km (through sites B and C), close to the foothills of the Castlecomer Plateau.

The lithological descriptions from the Clonmore feature are strikingly similar to those at the nearby unusual Neogene-aged deposit (Parkes et al. 2016) at Hollymount (Fig. 1), as described previously. If the trend of the Clonmore feature is extrapolated a further 2 km northwards, it reaches that Hollymount location. Palynological dating would be needed to correlate the feature to other areas of Neogene-aged fill, and further surface geophysics investigation would also be necessary to confirm whether the features are interconnected.

Pilot wells at sites B and C did not intersect the deepest low-resistivity anomalies, which extend beyond the maximum penetration depth on the 2D resistivity profile (100 m), equivalent to 40 m below present sea level. Clear evidence already exists for deep karstification of Irish Carboniferous limestones (e.g. Drew and Jones 2000; Schuler et al. 2018; Dolan et al. 2021), presumably formed during times of low relative sea level. Such instances are thought to have occurred during several periods throughout the Neogene–Pleistocene, with, for example, evidence from around the Irish coast supporting a sea level of about 80 m below present sea level in the Pleistocene (Plets et al. 2015).

The overlying low-permeability superficial deposits confined groundwater flow in the bedrock – drilling observations noted that the main groundwater ingress was at the top of bedrock, with the static level rising to within 2–3 m of ground level.

During the first (air flush rotary) phase of drilling, when borehole stability was harder to achieve, it was not always possible to place slotted screen intervals (Table 4) at the desired intervals. Because ground conditions were better understood after the first drilling phase, well design and construction improved for the second phase (drilling of PA-2 and its observation well). These improvements included: (1) resolving borehole stability issues by using mud rotary drilling; (2) achieving desired screen intervals; and (3) use of continuous wire-wound stainless steel (CWWSS) for the screen, which is of higher open area than the slotted PVC used in the first phase. The CWWSS screen used was a Johnson Hi-Flow™ 304 stainless steel screen.

Four single-well constant rate pumping tests (CRTs) were performed for periods of at least 3 days at rates varying from 375 to 1680 m³/day (Table 4). Owing to the increased open area of the screen material used, the specific capacity of PA-2 was much higher than that for PA-1 (even considering the increased losses from the higher pumping rate at PA-1).

Step tests were performed at both production wells to understand the relationship between the pumping rate and the drawdown (Table 5). Such tests also form a baseline for comparison against later operational performance.

Despite the anisotropic aquifer setting, use of the Theis equation for confined aquifers was considered suitable at a local scale. Using this equation, a best-fit transmissivity for Site A of 320 m²/day was obtained from a combined CRT for PA-1 and PA-2 carried out following the second phase of drilling (Table 6). Test results were considered to confirm the high groundwater development potential for Site A.

At sites B and C, considerably lower transmissivity values were achieved than at Site A (Table 4). In addition, at Site C barrier boundaries were encountered that led to continued drawdown in the well (an unstable rate of change of groundwater level), and pumping at the tested rate would have proved unsustainable over longer periods.

Elsewhere in Ireland, the presence of thick glacial till has greatly reduced the ability of effective rainfall to reach underlying bedrock aquifers, ultimately affecting the water supply (Misstear et al. 2008). With similar conditions here, identification of aquifer boundaries from the CRT data was important for potential groundwater source development. At Site A, observation well drawdowns during single-well CRTs at PA-1 and PA-2 (Fig. 4a) stabilized during the pumping tests (i.e. a recharge boundary was identified, with shallowing of the slope of the drawdown curve). Stabilization is typically attributable to induced leakage from an adjacent aquifer that eventually balances the abstraction rate, or induced recharge from a fully penetrating recharge boundary that eventually balances the abstraction rate (Misstear et al. 2006).

Image well analysis (Domenico and Schwarz 1997) was carried out, using the two single-well CRT results, to assess the location of the recharge boundary at Site A. For each test, a radius of distance to the recharge boundary was calculated (Fig. 4a) and resulted in two circles of radius 1006 m ((r_1.2) from OA-1) and radius 1244 m ((r_2.2) from OA-2). The intersections of these radii (Fig. 4b) are the two potential recharge source locations – the more likely of these is considered to be the westernmost as it is within about 100 m of the Fushoge River. Alluvial deposits of the river are thought to connect the surface water and deeper Site A groundwater flow system (Fig. 5). In the absence of flow information at the Fushoge River, an assessment of any river flow reduction as a result of groundwater abstraction is not possible. Other studies in limestone catchments have illustrated that low summer river flows may not be strongly influenced by groundwater abstraction, unless it is very close to the river (Rushton 2002).

Level monitoring at Site C pilot wells during the combined Site A pumping test did not change significantly, supporting evidence of a recharge boundary between the two sites. Of the pre-existing agricultural wells inventoried during Stage 1, only well 1 (see Fig. 4b) was measurably affected by pumping, suggesting that the cone of drawdown extends preferentially NNE along the trend of the fault zone. Inventoried wells 2 and 5 abstract water from the shallow superficial deposits and, although closer to the production wells, were not affected. This suggests that the regional upper sand and gravel aquifer is not significantly connected to the Site A groundwater flow system.

The results prompt a revision of the conceptual hydrogeological model for this part of the Barrow Valley (Fig. 5) to consider additional complexity in terms of dolomitization, karst features and recharge boundaries.

Intermittent groundwater level monitoring commenced at the three sites in mid-2008. Sites B and C were eliminated from the monitoring programme in 2009 but monitoring continued until August 2010 at Site A. In the study area, groundwater flows from NW to SE with a gradient of 0.003. Site A has the lowest potentiometric levels, at an average of about 50 m above sea level. The depth to groundwater was about 1–3.8 m below ground level over the monitoring period, falling during a period of low effective rainfall throughout 2010 (Fig. 6a).

Groundwater quality results from pilot wells at the three shortlisted sites (Table 7) indicate that the major ion chemistry is relatively similar across the three sites and these are calcium-bicarbonate waters – although magnesium is more elevated at Site A owing to the presence of dolomitized limestone, as observed elsewhere in Ireland (e.g. Tedd et al. 2017; O'Connell et al. 2022). Levels of nitrate were low across the three sites, suggesting that there was limited interaction with overlying or eastern sands and gravels (where moderate to high nitrate concentrations are known to occur: Table 3).

There were indications that elevated manganese was an issue across the study area, exceeding the threshold of 50 µg l⁻¹ (EU Directive 2020/2184) at sites B and C (the method detection limit for the Site A sample was higher than the threshold for manganese). High concentrations of iron were also present. Low levels of dissolved oxygen were found in the confined groundwater of Site A, a factor in elevated manganese in groundwater (e.g. Tedd et al. 2017; Kousa et al. 2021). Therefore, it is interpreted that reducing conditions play a part in the release of these elements from the bedrock sources to groundwater. Elevated manganese in groundwater from Carboniferous limestones has been identified in Ireland (Tedd et al. 2017) and Scotland (Homoncik et al. 2010), and the presence of the nearby Namurian shales has been associated with elevated manganese in similar settings (e.g. Derbyshire, UK: Abesser and Smedley 2008).

The key outcome from the previous feasibility and site survey work was that Site A had significantly greater hydrogeological potential for development as a water supply compared to the other two shortlisted sites. The LA commissioned the engineering design and construction of the surface infrastructure for a water supply scheme with a capacity of 1500 m³/day at Site A in 2011. The main water treatment consideration (in addition to standard chlorination and fluoridation) was for a reduction in the naturally elevated manganese levels. Manganese treatment is achieved via sorption onto a sand medium (glauconite/greensand), an effective and common treatment (Tobiason et al. 2016). No groundwater data collection took place during this stage.

Non-pumping groundwater levels (daily minima) were extracted from the hydrostatic probe telemetry data for January 2012–February 2014 (Fig. 6b). Although the pumping rates are set at 820 m³/day for PA-1 and 1300 m³/day for PA-2, pumps operate for about 60% of the time. This resulted in monthly average abstractions of between 1100 and 1400 m³/day (Fig. 6b).

Although the operational dataset is of a limited timespan (2 years), it captures a period during which the area experienced meteorological drought (defined in Ireland as 15 days with <0.2 mm rainfall) during July 2013 (Met Éireann 2013). More significantly, this was propagated to hydrological drought following the prolonged (almost 8 month) period of low effective rainfall from February to September 2013, with a groundwater recession (i.e. period of groundwater level fall) at the PA-1 hydrograph (Fig. 6b), during which non-pumping groundwater levels fell by about 3 m. The pumping water level at PA-1 during that time, although not shown, fell by a similar amount. The dataset records the period during which the drought was relieved; a subsequent 5 months during which more than 500 mm effective rainfall recharged the groundwater flow system and led to a recovery of groundwater levels of almost 4 m, to conditions higher than when the operation phase began in January 2012.

Raw groundwater quality during the operation stage was measured at both the production wells at Site A (Table 7) for fewer parameters than during the feasibility stage. Groundwater quality was similar between the two wells, and had a comparable specific electrical conductivity (EC) to that of the feasibility stage analysis. Manganese levels remained elevated above the drinking water standards at both wells, and nitrate was below the detection limit.

The low level of nitrate has continued from the feasibility stage through to the operation stage, indicating that induced leakage/recharge of higher-nitrate shallower groundwater has not occurred since activation of the water supply. During the wellfield operation period studied, the level of minerals within the groundwater (indicated by the specific electrical conductivity) were about 15% higher during the drought period (July 2013) compared to the winter period (February 2013), a pattern also noted during other seasonal studies of carbonate groundwater in Ireland and attributed to longer periods of water–rock interaction, mineral dissolution and accumulation of infiltrating solutes (O'Connell et al. 2022). Precipitation and seasonality has been found to facilitate microbial and contaminant transport within some Irish groundwater supplies (e.g. Fennell et al. 2021; O'Dwyer et al. 2021) but changes in the study area between the winter and summer periods related to microbial transport or nutrients were below measurement thresholds.

Assessing reliable yields using a graphical method (Misstear and Beeson 2000) requires the display of step-test information (Table 5), as well as information on monthly static and pumping (dynamic) water levels, to define an operational envelope of groundwater levels expected during a range of operational conditions. A ‘deepest advisable pumping water level’ (DAPWL) also needs to be set. Because the aquifer underlying Site A is confined, the DAPWLs for both wells there were set at the top of bedrock, according to standard practice (Sterrett 2007): 27 and 21 m below ground level for PA-1 and PA-2, respectively.

Pre-development step-test information was used to define the upper limit of the operational envelope. The important lower limit of the operational envelope, the ‘drought-bounding curve’, is set by plotting the lowest observed pumping water levels (i.e. from drought periods, as were experienced in 2013). Because PA-2 has a more efficient well construction than PA-1 (due to the use of high-open-area screen (Table 5)), the operational envelope extends to a much higher pumping rate than PA-1 (Fig. 7) before it intersects the DAPWL.

The intersection point of the DAPWL and the operational envelope is the reliable yield of the well (subject only to aquifer constraints): at PA-1, this corresponds to 1300 m³/day and at PA-2 it is 2800 m³/day (4100 m³/day total).

The difference in the operational envelopes illustrates the importance of construction in maximizing the well yield (Sterrett 2007): despite a more favourable hydrogeological position (closer to the recharge boundary), PA-1 is subject to greater well losses, so can only sustain about half the reliable yield of PA-2. It also illustrates that both wells currently operate significantly below their maximum reliable yields – pumping rates could be increased, if not subject to other constraints such as well diameter and pumping station storage capacity.

Standard analytical approaches for the zone of contribution, which rely on groundwater flow gradients and aquifer homogeneity, are ineffective in complex karst hydrogeological settings (Drew 2008) such as those encountered here. This assessment of operational groundwater level and quality is an adaptive management approach, a paradigm that has received increasing recent attention in groundwater management (Thomann et al. 2020). With this approach, key data emerge under periods of high meteorological stress (such as droughts) and recharge events, allowing interventions or ‘early warning’ levels to be developed and refined. Groundwater recharge events are expected to occur later in the year under climate change conditions (Morrissey et al. 2021), so adaptive planning for water supply will become increasingly advisable.

A new low-nitrate water supply from a groundwater source was required for a town in the Irish Midlands. An investigation comprising surface geophysics, pilot well drilling and long-term monitoring of groundwater level and quality at a study site revealed complex interactions between bedrock geological structure, karstification, and variable glacial and alluvial cover. The main outcomes of the study were three-fold, as detailed below.

The study identified new areas where Dinantian limestones were found to be highly dolomitized (Site A). Results indicated that the dolomitization is associated with a NNE-trending fault, and possibly associated with a body of regional dolomite mapped near the Leinster Granite, east of the Barrow Valley. The study led to the successful development of a new water supply for the town, with two wells (PA-1 and PA-2) abstracting from the dolomitized fault zone (Fig. 5). The more efficient well construction (screen open area) of well PA-2 results in it being able to produce over twice the flow of well PA-1. The groundwater source is low in nitrate, and confined conditions exist at the wellfield. The naturally elevated level of manganese requires treatment before transmission into the drinking water network.

Pumping test analyses at the wellfield (Site A) identified the likely source of recharge to be the Fushoge River, about 1 km away. Alluvial deposits of the river are thought to connect the surface water and the deeper groundwater flow system (Fig. 5). Water quality patterns did not appear to change significantly during the initial 18 months of operation, suggesting that the influence of shallower, higher-nitrate groundwater flow systems is limited. Despite the drought conditions of 2013, and an approximately 8 month period of groundwater recession, levels remained well above the deepest advisable pumping water levels at the wellfield, and recovered quickly during winter recharge events. An assessment of the reliable yield of the wells (subject only to aquifer constraints) concluded that they could supply up to 4100 m³/day total, compared to ongoing abstraction rates of about 1400 m³/day.

Elsewhere in the study area (sites B and C), a new large-scale infilled karst feature was identified. The feature (termed the Clonmore feature) trends north–south along the west of the study area, is likely to be continuous between sites (i.e. with a length of at least 3.5 km) and has a depth of over 100 m in places (Fig. 7). Lithological descriptions from the Clonmore feature are strikingly similar to those at a nearby known Neogene-aged deposit at Hollymount, and the two features may be linked.

This paper is dedicated to Eugene P. Daly, who developed the exploration strategy for the water supply, and provided direction, knowledge and advice throughout. SLR Consulting were the contracted engineering consultancy during the initial two stages of the project, and Minerex Geophysics performed the geophysics. Laois County Council and Irish Water made project information available for publication, and their staff, Paraic Joyce and Mark McCaulay, greatly supported the execution of the project. My thanks go to Connie O'Driscoll of Irish Water, Jafar Al-Jawad and two anonymous reviewers for their comments.

ODH: writing – original draft (lead).

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

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

All data generated or analysed during this study are included in this published article.