Use of electrical resistivity tomography to reveal the shallow freshwater–saline interface in The Fens coastal groundwater, eastern England (UK)

The Fens is a region that contributes 11% of the agri-food economy from just 4% of the agricultural land covering England (UK). This region is vulnerable to soil salinisation from sea-level rise with estimated 100-year flood events projected to be observed up to every 2 years by 2100. Seawater intrusion and upwelling of saline groundwater can provide an additional pathway; however, the area’s groundwater has not been assessed and the risk is unknown. This study used data from the British Geological Survey’s stratigraphic core archive to produce the first stratigraphic map of the loosely consolidated Holocene deposits in the South Holland–Holbeach Marsh region. There is a sandy unconfined aquifer towards the coast, a semiconfined central region with a silty cap and a clay confining cap in the north region. Electrical resistivity tomography data indicate water level depths of 0.58 ± 0.37 m above mean sea level (msl) in February 2021 and 0.01 ± 0.72 m msl in August 2021. The saline–freshwater boundary was at 1.70 ± 0.82 m msl in February 2021, deepening to 2.00 ± 1.02 m msl in August 2021, but the only evidence of seasonal fluctuation was within 10 km of the coast. A potential, but unverified, freshwater lens up to 3.25 m thick may exist beneath the surface. These results suggest that freshwater–saline interface fluctuations may primarily be driven by surface hydrology and would be vulnerable to climate-change-induced future variations in factors that affect surface water.


Salinisation in the United Kingdom (UK)
Salinisation of soil poses a serious threat for food production in many regions across Europe affecting around 30.7 Mha (Daliakopoulos et al. 2016;Gould et al. 2022;. This is particularly the case in the flat low-lying Holocene deposits of northern Europe-for example, in the coastal Netherlands, salt levels are projected to increase up to 140% as a result of saline seepage from sea level rise (Oude Essink et al. 2010). This increase in salt load may affect the salinity of the root zone, inducing soil salinisation.
Soil salinisation is not currently recognised as a major concern because there is significant rainfall in the winter coupled with low evaporation rates, allowing for any salt accumulation during the summer to be flushed in the winter (Cooper et al. 2010). However, coastal inundation is identified as a risk of soil salinisation for low-lying areas such as the Anglian coastal Fens, particularly since flooding is the UK's most serious natural hazard (Thorne 2014). The risk is especially high along the east coast that borders the North Sea characterised by extremely low-lying marshlands and wetlands (Daliakopoulos et al. 2016). The North Sea is well known for its storm surges, with 19 storm surge events of varying severity impacting the east coast of England between 1897 and 2007 along with a further devastating storm surge recorded in December 2013 (Spencer et al. 2015). These events have led to the inundation of agricultural land with seawater and thus resulted in the loss of crop yields. Flooding events on the east coast could have significant economic impacts and may affect up to 340,000 ha of cultivated crops at an estimated cost of around £5,000/ha, which could lead to longer-term yield penalties as saline soil recovers over multiple years (Gould et al. 2020).
Furthermore, under Representative Concentration Pathway 8.5 (RCP8.5), both mean sea level (MSL) rise and storm surge event frequencies are projected to increase more in the North Sea region than anywhere else across Europe's coastlines, leading to 100-year flood events, potentially being observed as frequently as every 20 years by 2050 or every 2 years by 2100 (Vousdoukas et al. 2017). Consequently, soils may not have time to fully recover from salinisation (Cooper et al. 2010).
Historically, groundwater research in the UK (England predominantly) has focussed on deep confined aquifers which typically are used for domestic water use (Barker et al. 1998;Hiscock et al. 1996;Mackay et al. 2015;Tellam et al. 1986). It is also well established that sea level rise has negligible effects on the placement of the freshwater-saline interface (Werner et al. 2012), whereas the impact on unconfined shallow coastal aquifers is debated (Ataie-Ashtiani et al. 1999;Kuan et al. 2012;Mao et al. 2006;Werner and Lockington 2006). Furthermore, the exploitation of unconfined shallow coastal aquifers for fresh groundwater resources can result in the lowering of the water table, sea-water intrusion, and upwelling of deep saline groundwater into shallow groundwater systems (Alfarrah and Walraevens 2018;Ferguson and Gleeson 2012;Klassen and Allen 2017). Currently, the authors are aware of only one other study in the UK, in Northern Ireland, that has surveyed and mapped shallow coastal groundwater systems using geophysical techniques which will inform evidence-based policy ensuring sustainable groundwater abstraction Águila et al. (2022).

UK Fens Agriculture
The Fens of eastern England contributes 11% of the agrifood economy from just 4% of England's agricultural land. Approximately, 50% of England's grade 1 (i.e. most productive) soils are located in this area producing a third of England's vegetables (NFU 2019). The high productivity of The Fens requires access to an abundant supply of water as 21.5% of all of England's intensive crops are grown in the region with an approximate value of £750 m (NFU 2019).
Climate projections in the UK show hotter and drier summers and warmer and wetter winters for the region (Lowe et al. 2018). The region of East Anglia has observed the greatest annual average temperature rise seen across the UK, increasing by more than 1 °C relative to the 1961-1990 average (Kendon et al. 2021), resulting in an increase in agricultural irrigation and is likely to be a major factor in why the water demand for irrigated cropping in South Holland-Holbeach Marsh area is amongst the highest in England reaching over 15,000 m 3 for every 2 km 2 (Knox et al. 2013).
The agricultural drought in 2018 highlighted the severity of the water stress in this area with many growers exhausting their licenced volume of water (Knox et al. 2020), and restrictions on agricultural irrigation were implemented in some areas (NFU 2019). In addition, climate change is predicted to increase the frequency and severity of droughts in the east of England adding further pressures to water resources. The source of most irrigation that takes place in The Fens is mains water from the public water supply (Water Resources East 2021) and in eastern England alone an extra 284 million litres of water per day will be required for agriculture by 2050 (Environment Agency 2020a); thus, The Fens are recognised as a region of increasing irrigation intensity (Weatherhead et al. 2014). The importance for adaption to the future climate was recognised for the South Holland-Holbeach Marsh area in the recent COP26 conference to ensure England does not become increasingly reliant on food imports (Water Resources East 2021).
Consequently, stakeholders in the South Holland-Holbeach Marsh area have started to gain interest in the possibility of abstracting fresh groundwater from the shallow aquifer beneath the coastal Fens as an alternative source of freshwater for agricultural irrigation; however, abstraction in 28% of groundwater bodies in the UK is currently unsustainable (Environment Agency 2018) and abstraction for irrigation is highest in eastern England (Knox et al. 2020). This may be detrimental to the quality of groundwater in lowlying coastal areas as unsustainable groundwater pumping has a greater and more rapid negative effect on groundwater salinisation than sea level rise (Ferguson and Gleeson 2012).
In the South Holland-Holbeach Marsh area, one method of abstraction is via catch-pits to fill storage reservoirs. If freshwater abstracted from the catch-pits was used for irrigation directly, an increased risk of water-table drawdown would result in a risk of saline seepage (Environment Agency 2020b). A further source of irrigation water is from surface-water abstraction; however, regarding farm water, testing finds these to be increasingly saline, with the likely saline source from subsurface waters. Landowners are permitted to abstract up to 20 m 3 of water per day without a licence; however, the amount of low-level abstraction that takes place is unknown. This is of particular concern as the groundwater system is unquantified in this area of The Fens as it has not been thoroughly assessed (Environment Agency 2020b; Rey et al. 2016).
Consequently, there is uncertainty about the threat of groundwater salinisation, and the sustainable management of groundwater systems cannot be achieved without the understanding of the stratigraphy or the geometry of the groundwater system. Therefore, given the importance of this area to the UK's agricultural productivity, it is of utmost importance to investigate the spatio-temporal distribution of the groundwater in the coastal shallow aquifer of the South Holland-Holbeach Marsh area.
This study aims to develop the first spatial interpretation of the risk of salinisation from groundwater in this region. The output will be the first maps of (1) the stratigraphy; (2) the water table; (3) freshwater-saline interface, and (4) the freshwater lens thickness of the South Holland-Holbeach Marsh coastal shallow groundwater system.

Topography, Geology, and Hydrology of Study Site
The study area is the coastal region of South Holland-Holbeach Marsh. Fenland soils comprise drained peatlands in Cambridgeshire, West Norfolk and parts of Lincolnshire, alongside Alluvial silt topsoils of South Lincolnshire. The location of the study is in the silty Anglian coastal Fens in the east of England bordering the North Sea and confined by the River Welland and the River Nene. In total it covers an area of around 35,000 ha which is predominantly used for agriculture with the main crop types being vegetables, potatoes, cereals and sugar beet (NFU 2008).
The Ordnance Survey's OS Terrain® 50 digital terrain model with a resolution of 50 m reveals that the South Holland-Holbeach Marsh topography is typically flat with a mean elevation of 2.25 m msl (Fig. 1). The highest point lies 7 m msl and is located in the centre of the study area northeast, whilst the lowest lying land is located in the south of the study area lying up to −2.10 m msl (Ordnance Survey 2021).
The soils are mostly calcareous, silty alluvial gleysols as classified by the World Reference Base system. Gleysols are derived from a range of unconsolidated parent materials including Holocene age marine sediments and can be found in temperate lowlands with shallow groundwater (IUSS Working Group WRB 2015). Once drained, gleysols can be cultivated for agriculture due to their fine soil texture and low rate of organic matter decomposition. As such, grade 1 soil, as classified by the then Ministry of Agriculture, Fisheries and Food (MAFF 1988), lies over the majority of the South Holland-Holbeach Marsh area (i.e. most productive). These soils allow for a range of crops to be grown with very little limitation to agriculture, with high yields typically of low variability in comparison to lesser quality soil (MAFF 1988), which is reflected in the higher value nature of cropping in the region (e.g. potatoes, salads, brassicas).
The superficial geology of The Fens is of Quaternaryage and mostly Holocene epoch marine alluvium (i.e. tidal flat deposits). These were formed through changes in sea level leading to multiple advances and retreats causing over 26 m of marine and estuarine clays, silts, and sands, now known as the Silt Fens, to be deposited (Natural England 2013;Sturt 2006). Based on typical hydraulic conductivities of unconsolidated sedimentary materials (Domenico and Schwartz 1997), this study considers the sand to be the shallow aquifer, the clay to be an aquitard and the silt to have low porosity.
The superficial deposits overlay the Jurassic-age bedrock geology of West Walton formation composed of Ampthill Clay and Kimmeridge Clay (undifferentiated), which are underlain by the (unproductive) Oxford Clay Formation which has negligible significance for water supply due to its low permeability, much lower than the aforementioned formations (Allen et al. 1997;Gallois and Cox 1976).
The South Holland-Holbeach Marsh area is located between the rivers Welland and Nene, which together drain an area of 395,000 ha. Weather data is collected and reported  (Parker et al. 1992;Alexander and Jones 2000). The mean annual temperature  for this region is 10.41 °C with a mean annual rainfall of 623.12 mm (Met Office 2021). In 2021, central England (which includes the east of England) had an annual rainfall of 660.8 mm with 248.90 mm falling in the winter (December, January, and February) and 146.50 mm falling in the summer (June, July, August) (Met Office 2021). The mean temperature for 2021 was 10.26 °C, with a mean temperature of 4.40 °C in the winter and 16.3 °C in the summer (Met Office 2021).
During the study, the region underwent periods of extreme weather. The combined total rainfall of December 2020 and January 2021 in the South Holland-Holbeach Marsh area was the wettest period on record. The Lincolnshire and Northamptonshire region had 246 and 223% increase in average rainfall for December 2020 and January 2021, respectively (Environment Agency 2021b, d), which was compounded by low evaporation rates due to an average temperature of 3.1 °C in January (Met Office 2021). By comparison, the rainfall observed during July and August in the South Holland-Holbeach Marsh area had less deviation from the long-term average, 105 and 43%, respectively (Environment Agency 2021a, e); furthermore, July was the hottest month of the year with an average temperature of 17.7 °C (Met Office 2021).
The British Oceanographic Data Centre does not have a tidal gauging station for The Wash (the 62,046-ha rectangular bay consisting of multiple estuaries which forms the coast of The Fens, see Fig. 1). Therefore, to understand the tidal fluctuations in The Wash tide heights at the two nearest tidal gauges, at the towns of Cromer and Immingham, were used. Cromer is located around 70 km east of The Wash and Immingham is located around 85 km north of The Wash at the River Humber estuary. At Cromer the tide height ranged from -0.76-5.34 m in the winter with a mean of 2.75 ± 1.29 m, whilst in the summer the tide height ranged from 0.52-5.28 m with a mean of 2.94 ± 1.18 m. At Immingham the tide height ranged from 0.45-7.58 m in the winter with a mean of 4.25 ± 1.71 m, whilst in the summer the tide height ranged from 0.72-7.49 m with a mean of 4.27 ± 1.68 m (British Oceanographic Data Centre 2021).

Mapping Aquifer Stratigraphy
To map the aquifer's stratigraphy, a reconstruction process was completed by compiling 122 previously logged British Geological Survey (BGS) stratigraphic core records (BGS 2020)-see section A of the electronic supplementary material ESM1-for the compiled record. The oldest stratigraphic core was taken 136 years ago in 1885, whilst the latest stratigraphic core was taken 9 years prior to the study in 2012 and 95% of the cores were collected and logged between 1938 and 2002. The historical stratigraphic cores were also used to help infer the ERT results later in this study and, as a result, the stratigraphic cores that were not of a length to make contact with the water level were removed from this study. After an initial review of all stratigraphic cores, the substrata were divided into three sedimentary categories-clay, silt, or sand. Out of the 122 stratigraphic cores, 78 also included the depth at which the water level (m bgl) was met. The stratigraphy was modelled using BGS Groundhog 2.5 software (BGS 2021) with a 50 m × 50 m digital elevation model supplied from the Ordnance Survey (Ordnance Survey 2021). The Nearest Neighbour interpretation method, built into Groundhog 2.5, was used to create the prediction surfaces for each layer of the stratigraphy. These prediction surfaces were then exported as raster layers. Finally, using ArcGIS 10.8, the prediction surfaces were clipped to fit the study area with the National Characters Area shapefile published by Natural England (Natural England 2021) and layered on top of each other to create the geological map (Fig. 3).

Geophysical Techniques to Map Shallow Groundwater Systems
One-dimensional (1D) geophysical techniques such as vertical electrical sounding and the profiling method have been used in the past to provide resistivity surveys of the subsurface century (Dahlin 2001;Gish and Rooney 1925;Koefoed 1979). However, these 1D electrical survey techniques produce inferred results with limited accuracy as they assume that the stratigraphy of the subsurface is homogenous in either the vertical or horizontal plane (Loke 1999).
More recently, two-dimensional (2D) electrical resistivity tomography (ERT) has become an increasingly popular method of surveying the groundwater in shallow coastal aquifers around the world (Aladejana et al. 2020;Águila et al. 2022;Hermans and Paepen 2020). Due to the difference in typical resistivity that characterises fresh groundwater (10-100 Ω) and seawater (0.2 Ω), the water level and freshwater-saline interfaces in shallow aquifers in coastal areas can be accurately mapped (Loke 1999). Furthermore, this method is noninvasive and thus is suitable for all soil types and is relatively cost-effective in comparison to other population techniques such as cone penetration tests (e.g. Pauw et al. 2017).
The resistivity results of 2D ERT are dependent upon both the geological properties of the coastal shallow aquifer and the salinity of the groundwater. As a result, it is frequently found that geophysical data cannot clearly distinguish between the interval changes in saturation and salinity in the groundwater (i.e the water level and freshwater-saline interface) from geological features, especially in low resistivity superficial geology (Szalai et al. 2009). Therefore, calibration of inferred water levels and freshwater-saline interfaces is required through corroborating information from hydrogeological data; one of the favoured methods of calibration is through stratigraphic core (lithological log) records (Galazoulas et al. 2015). Therefore, using an integrated approach with geophysical results and geotechnical information used in tandem the certainty and robustness of the inferred groundwater level is higher.

Site Selection and Geophysical Analysis
The first fieldwork campaign took place between the 1st and 12th of February 2021 and the second fieldwork campaign was carried out between the 24th and 29th of August 2021 to allow for maximum seasonal change to occur. Historically  in Holbeach February is the coldest and second driest month of the year with an average minimum temperature of 1.70 °C and an average rainfall of 38.53 mm and August is the warmest and wettest month in Holbeach with an average maximum temperature of 21.78 °C and an average rainfall of 64.18 mm (Met Office 2021). Two-dimensional electrical resistivity tomography (ERT) was used to survey the water level and freshwater-saline interface in the South Holland-Holbeach Marsh shallow aquifer. ERT is a wellestablished method that measures the resistivity distribution in ohm-metres (Ωm) across predetermined survey lines allowing for an indirect measure (proxy) of subsurface hydrogeological heterogeneities such as changes in stratigraphy or groundwater and is also a well-established method to assess freshwater-saline interfaces in coastal aquifers (Loke 1999).
The Allied Tigre earth resistance metre was used with a 32-electrode cable to survey the groundwater in the subsurface, whereby 27 sites were selected (Fig. 2) mostly adjacent to existing BGS cores, which allowed for single borehole-to-surface ERT measurements, whilst maintaining a good spatial spread covering the study area as best as possible. A cross-hole configuration was not used as the stratigraphic core separation/depth ratio was greater than 1, which would have produced inaccurate results (Tsourlos et al. 2011). The electrodes were spaced 2.5 m apart, allowing for good resolution and a satisfactory investigatory depth of 12 m. The total length of the survey line was 77.5 m. Each electrode was secured to the soil beneath with a steel peg, and resistance checks were carried out with software RES2DINV to ensure contact resistances were not inordinate. The processing software used in this study, RES2DINV, allows for changes in topography across long survey lines to be accounted for in the data processing, and due to the extremely flat nature of the South Holland-Holbeach Marsh area, it was determined that the slight changes in elevation between each electrode would be negligible in regard to survey quality. The survey lines were, where possible, orientated perpendicular relative to the local coastline (NE-SW). The surveys obtained subsurface resistivity measurements using the Wenner (α) array. This array was selected over other commonly used electrode arrays in environmental geophysics (Wenner-Schlumberger and Dipole-Dipole) due to its sensitivity to vertical changes in the subsurface whilst maintaining a good investigatory depth (Barker 1979;Loke 1999;Reynolds 2011).
The raw resistivity results, processed in RES2DINV, showed the subsurface strata resistivity variations for each survey site. These were checked with information from the historical stratigraphic core records previously collated in this study to cross-check and identify possible changes in geology or ground condition which may have needed to have been accounted for. The field area was selected specifically to ensure that geological, geomorphological, and hydrological factors were likely to be as consistent as possible. Then each ERT profile was processed and analysed individually on its own merits and then these profiles were brought together and assessed collectively to establish both local/ regional links and patterns as well as to identify outliers in the observed data.
To determine the unsaturated, semisaturated, saturated (freshwater), and saline interface, distinctive rates of change in resistivity were required that were broadly observed and reproducible across the area. Upon analysis of all the ERT survey data, these values were determined to be of the following characteristic trend: ×1.25-1.75 apparent resistivity decrease inferred the change between unsaturated zone (near surface soils) and the semisaturated and saturated (freshwater) zone, and a ×4-×5 decrease inferred the depth of the interface between fresh and saline water. Apparent resistivity is reported as it is a weighted average of resistivity across the survey lines, which helps smooth out localised variations that are poorly understood given the poorly studied nature of the region.

Interpolating Electrical Resistivity Tomography Results using Geostatistics
The water level and freshwater-saline interface was inferred from the processed ERT data (see section 'Mapping aquifer stratigraphy'). To create a prediction layer for the study area, R studio was used, which utilises different packages for geospatial analysis. To spatially interpolate the point data, the "autoKrige" function (part of the automap package (Hiemstra 2015), which executes automatic kriging on the data points collected at each ERT survey site, was also used. The autokrige function automatically generates an omnidirectional variogram and fits it to 1 of 3 models (spherical, exponential or gaussian), after which a prediction surface is outputted at a 50 m × 50 m resolution. This process was repeated for the winter water level and freshwater-saline interface and summer water level and freshwater-saline interface. Cross-validation was also carried out during the geostatistical analysis using the 'autokrige.cv' function in the 'automap' library to ensure as little error as possible. Each prediction surface was then exported as a raster layer and imported into ArcGIS. The raster layers were clipped to fit the study area and the final figures were produced. A freshwater lens thickness map for both seasons was also produced by calculating the difference between the water level depth and freshwater-saline interface depth at each survey site (freshwater lens thickness = freshwater-saline interface -water level).

Electrical Resistivity Tomography Validation
To validate the ERT results, the water level depth inferred from the ERT surveys was compared to measured water level depths documented on past stratigraphic cores located next to the ERT survey sites. Of the 27 ERT survey sites, 10 had adjacent stratigraphic cores with documented water levels and ages ranging from 1963 to 2012 with a mean of 1,976 ± 4.71 years. During this time the water level in The Fens remained relatively stable (Dawson et al. 2010) which allowed this data to be sufficient to validate this study's ERT results against.
A lack of fit test (Whitmore 1991), Pearson's correlation, and root-mean-square-error (RMSE) were analysed. All analysis was performed in R using the basic functions along with the alr4, and Metrics package for Pearson's correlation, lack of fit test, and RMSE, respectively.

Stratigraphy
The South Holland-Holbeach Marsh stratigraphy follows a layer cake formation with three facies; on top is a clay cap, followed by a middle layer of silt and a bottom layer of sand (Fig. 3). The clay cap is mostly present in the north of the region and extends southwest inland adjacent to the River Welland. The silt facies outcrops across the central area of the study area extending to the River Nene in the southeast. Within the silt outcrop relatively small clay caps are found such as the one that the Market Town of Holbeach is situated on. The sand facies outcrops at the coast in the northeast of the study area and broadens into Long Sutton adjacent to the River Nene. The sand is observed to outcrop again around 18 km inland southwest from the coast.

Electrical Resistivity Tomography
The following section describes the general pattern of the raw ERT data for all 27 ERT surveyed sites. Please see the supplementary information for the raw ERT survey files in ESM2, and see ESM1 for each individual ERT survey output.

Winter Survey raw ERT Results
For the winter survey, the apparent resistivity of the upper layers (1.5-2.5 m bgl) was mostly characterised by a relatively high apparent resistivity between 50 and 175 Ωm (typically >75 Ωm), which indicates that soils are restricted to depth of no more than 1.5-2.5 m bgl (upper horizon) and was inferred as the unsaturated zone within the stratum. Below this, typically between 1.5 and 3.0 m bgl, the apparent resistivity was lower, mostly between 8 and 50 Ωm (typically <45 Ωm), but gradually decreasing with depth. This was interpreted as the partially saturated and uppersaturated zone of the water level between depths of 1.5 and 3.0 m bgl. Further down the apparent resistivity decreased exponentially with depth below ~2.5 and 3.5 m bgl with the characteristic electrical resistance being 4-5 times smaller in the lower saturated zone than the upper saturated zone.

Summer Survey raw ERT Results
In the summer, the apparent resistivity indicated that the soils at depths between 0.5-1.5 m were characterised by relatively high apparent resistivity (typically >75 Ωm). The soils between 0.5-1.5 m bgl were therefore inferred as the unsaturated zone within the stratum. Below this, between 1.5-3.0 m bgl, the typical apparent resistivity was lower between 18 and 65 Ωm (typically <45 Ωm) and gradually decreasing with depth. The soil between 1.5-3.0 m bgl was interpreted as the partially saturated and upper-saturated zone of the water level. Further down, the apparent resistivity decreased exponentially below depths of around 2.5-3.75 m bgl with the characteristic electrical resistance being 4-5× smaller in the lower saturated zone than the upper saturated zone.
The apparent resistivities of a representative coastal survey site (MM02) and a representative inland site (MM23) Fig. 3 Geological map of the South Holland-Holbeach Marsh study area with two cross sections provided. White-shaded area shows no data were available for that region are shown in Fig. 4c,d respectively. The coastal MM02 has apparent resistivity values of 18-22 Ωm, 5.4-6.5 Ωm, and 2.8-4 Ωm at depths of 4, 8, and 12 m bgl, respectively, whilst the inland MM23 indicated apparent resistivity values of 22-25 Ωm, 6.25-8.5 Ωm, 3.5-5.25 Ωm, respectively.

ERT inferred winter groundwater
The inferred water level ranged from -1.65-2.10 m msl. The deepest water level was found around 18 km inland at MM23, whilst the shallowest value was observed around 7 km inland adjacent to the River Nene at MM13. The mean depth of the winter water level was 0.52 ± 0.91 m msl.
The inferred freshwater-saline interface ranged from -3.65-0.73 m msl. The deepest freshwater-saline interface was found at MM23, whilst the shallowest value was located around 1 km inland at the coastal site MM01. The mean depth of the winter freshwater-saline interface was-1.44 ± 1.12 m msl.
The freshwater lens thickness ranged from 0.63-3.25 m. The deepest value of freshwater was found around 14.5 km inland adjacent to the River Nene at MM19a, whilst the lowest value was found at MM01. The mean freshwater lens thickness in the winter was 1.95 ± 0.67 m.

ERT Inferred Summer Groundwater
The water level ranged from -2.65-1.98 m msl. The deepest water level was found around 11.3 km inland at MM29, whilst the shallowest value was found towards the coast around 2.5 km inland at MM03. The mean depth of the summer water level was 0.25 ± 1.17 m msl.
The freshwater-saline interface ranged from -4.65-0.78 m msl. The deepest freshwater-saline interface was located around 18 km inland adjacent the River Wellend at MM25, whilst the shallowest value was found towards the coast at MM03. The mean depth of the summer freshwater-saline interface was -1.55 ± 1.40 m msl. The freshwater lens thickness ranged from 0.75-3.13 m. The deepest value of freshwater was found at MM23, whilst the shallowest value was located at MM29. The mean freshwater lens thickness in the summer was 1.80 ± 0.59 m.

Winter Groundwater Geostatistical Results
The variograms and a description of the variance are provided in section D of ESM1; however, for the winter, the water level spherical model was fitted with a nugget of 0.04, a sill of 0.86, and a range of 4,070. The freshwater-saline interface fitted the Gaussian model with a nugget of 0.45, a sill of 1.71, and a range of 8,040, whereas the variogram for the winter freshwater lens thickness fitted the Gaussian model with a nugget of 0.00, a sill of 0.31, and a range of 2,790.
The water level ranged from -1.52-2.02 m msl (Fig. 5a). The deepest water level was found at MM23, whilst the shallowest was located at MM13. The mean winter water level was 0.58 ± 0.37 m msl. The error variance ranged from 0.09-0.95 m and had a mean of 0.79 ± 0.29 (see section E of ESM1 for variance error maps).
The freshwater-saline interface ranged from -3.12-0.64 m msl (Fig. 5b). The deepest freshwater-saline interface was located around 3 km south of MM23, whilst the shallowest value was found at the coastline around 2 km northeast of MM02. The mean winter freshwater-saline interface depth was -1.70 ± 0.82 m msl and follows a relatively consistent spatial distribution becoming deeper the further southwest (inland) the groundwater is located. The error variance ranged from 0.54-2.55 m and had a mean of 1.01 ± 0.54 m.
The freshwater lens thickness ranged from 1.04-2.92 m (Fig. 5c). The freshwater lens was located 0.1 km east of MM19a, whilst the thinnest freshwater lens was located at the coast 0.5 km northwest of MM01. The mean freshwater lens thickness in the winter was 1.95 ± 0.27 m.

Summer Groundwater Geostatistical Results
The variogram for the summer water level was fitted with the exponential model with a nugget of 0.48, a sill of 3.87, and a range of 35,400. The summer freshwater-saline interface fitted the exponential model with a nugget of 0.73, a sill of 5.86, and a range of 61,700. The freshwater lens thickness was fitted with a Gaussian model with a nugget of 0.14, a sill of 0.26, and a range of 2,820.
The water level ranged from -1.44-1.30 m msl (Fig. 5d). The deepest water level was found at MM29, whilst the shallowest was located at the coast around 1 km north of MM01. The mean summer water level was 0.01 ± 0.72 m msl. The error variance had a range of 0.65-2.15 m and a mean of 1.03 ± 0.29 m.
The freshwater-saline interface ranged from -3.75-0.14 m msl (Fig. 5e). The deepest freshwater-saline interface was located at MM25, whilst the shallowest was found at the coast around 2.5 km north-east of MM02. The mean summer freshwater-saline interface depth was -2.00 ± 1.02 m msl. Similar to the winter study, the freshwater-saline interface follows a relatively consistent spatial distribution becoming deeper the further inland the groundwater in located. The error variance ranged from 0.94-2.38 and had a mean of 1.29 ± 0.28.
The freshwater lens thickness ranged from 0.69-3.13 m (Fig. 5f). The thickest freshwater lens was located at MM23, whilst the thinnest values were found around 0.6 km south of MM29. The mean freshwater lens thickness in the summer was 1.87 ± 0.42 m.

Seasonal changes
Overall, across the whole study area there was no statistical difference between either the winter and summer water level (p = 0.68, n = 25) or freshwater-saline interface (p = 0.80, n = 25). Of the 25 sites measured in both the summer and winter, 12 sites had higher water levels and 11 had lower levels, whilst 12 sites had a higher saline interface and 12 had a lower saline interface. However, there is a significant positive correlation between the change in water level between summer and winter and the change in saline boundary between summer and winter (Fig. 6a, r = 0.78, p < 0.01, n = 25). Two outliers for this were MM19 and M23 both of which are two of the most inland sites.
It was only amongst the near coastal sites from MM01 to MM11 which are mostly <10 km from the shore that there was a significant rising in the saline interface from winter to summer. This was an average decrease of 30 cm between the seasons (p = 0.04, n = 11) no corresponding change was observed in the water level (p = 0.20, n = 11).

ERT Validation
The statistical analysis suggests that the inferred ERT provide a reliable representation of the groundwater level in the South Holland-Holbeach Marsh area (Fig. 6). The lack of a fit test shows a high relationship between the inferred and measured water levels (p < 0.001). Pearson's correlation coefficient between the inferred and measured water level was significant (r = 0.96, p < 0.001, n = 10, Fig. 5b), while the RMSE between the inferred and measured water level was 0.69 (n = 10).
Cross validation was used to ensure error in the kriging data was as small as possible. The most robust prediction (kriging) models have mean errors (ME) that should be 0, the RMSE should be as small as possible (the autokrige function provided this, fitting each variogram to the best fitting model) and the root mean square standardized (RMSS) should be close to 1 (Adhikary et al. 2014). Results from the cross validation (Table 1) show that the ME is extremely low making it almost negligible, the RMSE ranged from 0.60 to 1.42 across all kriging predictions, and all RMSS values are close to 1, suggesting satisfactory robustness of the kriging predictions reported in this study.

Potential Influence of Stratigraphy on Salinisation
The stratigraphic survey undertaken by this study (Fig. 2) of the South Holland-Holbeach Marsh area suggests that direct groundwater recharge is most likely to be restricted to the coastal system and extends southwards adjacent to the River Nene towards Long Sutton. This is due to a sand outcrop in this region which is characteristic of an unconfined aquifer (Domenico and Schwartz 1997). The central region of the study area consisting of a silt outcrop may act as a semiconfining layer. A confining clay cap is located in the north-westerly region and extends along the length of the River Wellend. Theoretically, this should provide an impermeable confining layer (aquitard) which typically has a lower hydraulic conductivity than the silt and sand outcrops (Domenico and Schwartz 1997).
In the coastal areas of The North Sea region the flushing of salts from soil is primarily driven by rainfall, and the rate of salt flushing is higher in more permeable soils (Christensen 2022). Therefore, depending on the length of seawater inundation the well-drained sandy soils in the South Holland-Holbeach Marsh area may also be able to recover back to preinundation conditions in 2 years significantly reducing the crop yield loss and the financial cost of seawater inundation (Gould et al. 2020).
However, in a soil with low permeability such as the clay cap present in this study area, the vertical movement of saline water after a flood event can be significantly slower. It may take up to 7 years for potentially poorly drained soils located in the region of the confined clay cap to recover from a seawater flood event significantly increasing the financial loss to growers in this region (Gould et al. 2020). In a comparable aquifer in the Netherlands remediation of sodium polluted water through flushing is considered to take up to 40 years (Appelo and Postma 2005).
The flat topography of the South Holland-Holbeach marsh area coupled with the low infiltration rate of the clay cap and the agricultural land use type also makes this area high risk to sodification (Tomaz et al. 2020). This may lead to the structural decline of soil and thus force a change in land management or permanent abandonment of agricultural practice in this region (Gould et al. 2020). Therefore, if seawater inundation was to occur beneath a clay cap in the South Holland-Holbeach Marsh area, both groundwater abstraction and agricultural practice in this region may be unsuitable for multiple years due to salt contamination.
Unconfined coastal regions are the most at risk from groundwater flooding (Upton and Jackson 2011), which may lead to seepage salinisation. Although sandy soils allow good direct recharge, eastern England (shown in Fig. 1), in which the South Holland-Holbeach Marsh area is located, receives the lowest amount of rainfall in the UK (Mayes and Wheeler 2013), and it is projected for summers to become drier (Lowe et al. 2018). In a comparable coastal aquifer in the Netherlands, saline seepage has been predicted to occur by the end of the century as a result of the freshwater-saline interface rising towards the surface (Oude Essink et al. 2010). Furthermore, the rate at which the salinisation takes place is accelerated in low-lying drained land due to land subsidence (Pauw et al. 2012). Therefore, the coastal system of the South Holland-Holbeach Marsh, which is mainly drained land, may likely be vulnerable to saline seepage when considering the permeably sand outcrop at the surface.
There are, however, caveats to these findings, particularly in the northern area where the clay cap and silt outcrop boundary is present. Here there is a lack of stratigraphic cores and geological data is scarce and, as such, the boundary is poorly constrained.

Fluctuations in Water Level and Freshwater-Saline Interface
The data indicates that for this shallow, low-lying Holocene coastal UK aquifer, there is no statistically significant fluctuation in the groundwater level. In commonly monitored hard rock UK aquifers, this is unusual since summer groundwater typically has lower levels than winter due to the high winter precipitation (Mackay et al. 2015). The high-water level might be sustained by connectivity to the seawater which fluctuates little for the site (British Oceanographic Data Centre 2021). Thus, the relatively constant sea level has the effect of maintaining a constant pressure head in the groundwater (Kuan et al. 2012), particularly since groundwater dynamics at this site are not topographically driven (Werner et al. 2012). There are, however, significant fluctuations in the salineboundary: this is probably primarily driven by surface hydrology. The saline boundary lowers from a summer high to a winter low within ca. 10 km of the coast. During the winter months, when the saline boundary was low, other systems including the neighbouring Welland and Nene rivers had water levels classified as 'extremely high'; however, by the summer, when the saline boundary was high, these systems were classified as 'below normal' (Environment Agency 2021c). Therefore, it is likely that winter freshwater is pushing the saline front downwards during these months (Fig. 7); however, to test this hypothesis further, intensive year-round monitoring of the groundwater system would need to be undertaken.
Nevertheless, should the surface water be controlling the saline front within the groundwater system, there would be consequences for this site and other similar systems. This is because, firstly, climate change is projected to induce increasingly unpredictable variations in rainfall (Taylor et al. 2013), and where this increases droughts, then aquifers like this will have a higher saline front than they currently do. Secondly, humans are projected to increase the use of water resources of both ground and surface waters-e.g. damming irrigation, etc. (Haddeland et al. 2014;Vörösmarty et al. 2010)-including at the locality where large reservoirs are planned (Anglian Water 2022). Where this reduces the availability of surface water, this may also lead to a higher saline front by reducing the downward flux of freshwater. For this to be confirmed, further intensive data is required from this site and other similar sites; however, in the UK, most shallow, low-lying alluvial coastal aquifers are not routinely monitored.

Potential for Freshwater Abstraction
Changes in weather patterns and rising sea levels are putting pressures on global water supplies for agriculture and irrigation, and eastern England is no different. Shallow unconfined groundwater systems have historically not been considered suitable for such irrigation due to an insufficient volume of freshwater availability to support groundwater abstractions. Anecdotally, growers in the South Holland-Holbeach Marsh area have typically viewed the groundwater salinity levels as being too high for effective agricultural irrigation. However, there has been increased interest in alternative freshwater sources not least because of the extra 284 million litres of water per day that will be required for agriculture by 2050 (Environment Agency 2020a). As a result, a Holbeach Marsh Abstraction Licensing Strategy has been produced with the possibility for groundwater abstraction through the installation of shallow catch-pits (Environment Agency 2020b).
The results show that up to 3.25 m of freshwater may exist in some, particularly inland, locations where the aquifer is confined. Combined with recent developments in saline farming , this indicates that this shallow system, and others of similar characteristics around the world, could potentially be utilised as and when a groundwater resource is required. This should be investigated further as a partial solution to the area's water needs; however, before widespread exploration and extraction takes place, a greater understanding of both the chemistry and hydrogeological dynamics must be gained. Understanding the hydrogeological dynamics is particularly important to understand the affect that pumping could have on the salinisation pathway from the sea.
The unconfined coastal aquifer (Fig. 2) is likely to have most vulnerability to saline upwelling through the permeable sand. In the future there may be increased influence from tidal forcing in The Wash that may potentially increase the amount of seawater intrusion, increasing the vulnerability to saline upwelling in coastal areas if groundwater abstraction was to take place here (Ataie-Ashtiani et al. 1999;Pauw et al. 2012).

Conclusion
This study has produced the first inferred assessment of freshwater-saline interface and groundwater level of the shallow coastal aquifer of the Anglian Fens, UK, and it is coupled to an interpretation of the historical BGS stratigraphic core records enabling the construction of a stratigraphic map of the South Holland-Holbeach Marsh area. This reveals that the aquifer is unconfined at the coast as a result of the sand outcrop but in the north of the study area and adjacent to the River Wellend, a clay cap is present resulting in the aquifer being confined. In the central region, the shallow aquifer is likely to be semiconfined due to the silt outcrop present. The ERT results from this study indicate the depth of the water level and freshwater-saline interface are very shallow. During both winter and summer seasons, the inferred freshwater-saline interface was closest to the surface near the coast and became deeper the further inland. Only in the coastal region, <10 km from the shore, do the tides appear to influence seasonal fluctuations in the depth of the saline water and such fluctuations appear to be in lockstep with the water level. Nevertheless, the work shows that in some locations up to 3.25 m of freshwater may exist beneath the surface, but more investigation of the chemistry and hydrogeology must be done to assess suitability for extraction.