1 Introduction

Soils are known to be very heterogeneous due to the variability in soil properties or soil taxonomic classes within an area (Maestre and Cortina 2002; McBratney and Minasny 2007), resulting from regional soil formation factors such as topography, parent material, climate, organisms, and time (ODNR 2018; Sposito 2023). At land-lake interfaces, or more generally terrestrial-aquatic interfaces (TAIs), heterogeneity in soil architecture is much more diverse. This is because the TAI is where terrestrial and aquatic ecosystems meet and interact, forming an active and dynamic zone where various hydrological and biogeochemical exchanges occur at different spatial and temporal scales, thereby introducing additional sources of heterogeneity in the TAI soils. The diverse heterogeneities embodied by coastal TAIs are usually not accounted for in current earth system models (ESMs) (Ward et al. 2020).

Soil architecture, which refers to the close relationship between the arrangement of soil physical components in space and the functioning that such arrangement enables (Baveye et al. 2018; Vogel et al. 2022), is controlled by the spatial configuration of pore networks resulting from processes of root growth, wetting and drying dynamics, freeze-thawing cycles, and tillage operations (Dexter 1988; Vogel et al. 2022). Other factors that control soil architecture include the metabolic activities of soil micro and macro fauna within a soil matrix, the cementing organic molecules, and associated physicochemical exchanges (Dexter 1988; Vogel et al. 2022). Soil architecture thus serves as a complex heterogeneous biogeochemical interface that forms the basis for various soil functions such as water retention, root growth, nutrient cycling, carbon storage, functional biodiversity, solute transport, and contaminant degradation (Totsche et al. 2010; Vogel et al. 2022). The extent to which these factors will influence the soil architecture depends on the soil type as well as the characteristics of the ecosystem unit or study site. Although these critical processes that control soil architecture occur mostly at the pore scale, their effects extend to larger spatial scales (e.g., site to regional scales), as many hydrological and ecological soil functions are governed by the soil architecture (Stewart et al. 1990; Romero-Ruiz et al. 2019).

Soil architecture has been investigated either by the aggregate approach or the pore approach. The aggregate approach targets the stability and composition of isolated solid fragments, while the pore approach targets the pore structure as well as the pore-solid interfaces in undisturbed samples (Rabot et al. 2018; Vogel et al. 2022). The aggregate approach is challenged by the limited understanding of how matter and energy fluxes through the soil will be affected when isolated from the original soil matrix. It is expected that fluxes of liquid, gas, or nutrients will differ between an isolated soil volume compared to an undisturbed one (e.g., Kravchenko et al. 2019). The pore approach considers the importance of spatial position but just within the context of an undisturbed sample. Although the pore approach recognized that flow and mixing processes such as diffusion of dissolved organic carbon, bioturbation, and pore water dynamics create spatial heterogeneity in soil architecture, it does not account for such heterogeneity beyond limited core samples (e.g., Young et al. 2001).

Traditional methods of soil investigation, such as soil cores, hand augers, excavation, or sensors (e.g., Osborne and DeLaune 2013), are point measurements that lack spatial resolution and may not adequately capture the spatial variabilities necessary to upscale models from site to global scale. Geophysical methods offer non-invasive techniques to capture spatial variability in soil properties at high resolution and across various spatiotemporal scales (e.g., Besson et al. 2013; Krüger et al. 2013; Emmanuel et al. 2023). Romero-Ruiz et al. (2018) reviewed the potential of harnessing geophysical techniques for the characterization of soil architecture and identified geoelectrical and electromagnetic methods among a spectrum of geophysical methods as ideal for soil architecture characterization. Due to their sensitivity to soil hydrological states, these methods, such as electromagnetic imaging (EMI) (e.g., Corwin and Lesch 2005; Brechet et al. 2012; Doolittle and Brevik 2014; Emmanuel et al. 2023), electrical resistivity tomography (ERT) (e.g., Michot et al. 2003; Kizhlo and Kanbergs 2009; Besson et al. 2004, 2013; Doro et al. 2013), induced polarization (IP) (e.g., Kemna et al. 2012; Kessouri et al. 2019), and ground penetrating radar (GPR) (e.g., Grote et al. 2003; Krüger et al. 2013), have the capacity to assess the soil pore space and how its varied distributions will affect soil hydrology.

Although geophysical methods have been used to provide high-resolution understanding of soil spatiotemporal variabilities, most geophysical investigations of soil are mainly focused on terrestrial ecosystems, thus the suitability of geophysical methods to map the diverse heterogeneities in soils across land-lake interfaces is not well understood. We hypothesize that the varying physicochemical properties of soils at the land-lake interface and its dynamic processes create measurable contrast in geophysical properties such as electrical conductivity and dielectric permittivity. Hence, geophysical methods could be useful to reveal the soil architecture and subsurface stratigraphic heterogeneities at a higher resolution than point sampling methods. We expect soil electrical conductivity to be high in the wetland zones due to wetting from the adjacent lake, as well as nutrient accumulation. As we move towards the upland, the soil gets dryer due to increased root water uptake and higher evapotranspiration, and soil electrical conductivity is expected to decrease. Based on this hypothesis, the objectives of this study were to (1) evaluate the sensitivity of geophysical methods to spatiotemporal variability of soil properties across land-lake interfaces, (2) investigate how the soil hydraulic properties will change from the uplands (coastal forest) to the wetlands (marsh), (3) test the suitability of different geophysical methods for reconstructing the subsurface stratigraphy of land-lake interfaces, and (4) evaluate the pros and cons of combining multiple geophysical methods compared to a single method.

In this study, we combined three geophysical methods (EMI, ERT, and GPR) with borehole information, as well as soil sensor and groundwater data, to characterize soil architectural properties and subsurface stratigraphic heterogeneities across land-lake interfaces along Lake Erie. We expect that combining the three geophysical methods will be useful to overcome the weaknesses of each individual method, enabling us to better understand both the lateral (using EMI) and vertical (using ERT and GPR) variations in soil properties across the TAI. This approach is expected to provide a non-invasive and detailed characterization of TAI soils at high spatiotemporal resolution, which is usually lacking with point sampling approaches. This could also guide direct sampling and monitoring campaigns, replacing haphazard sampling and providing essential data for constraining pedophysical and hydrological models across the TAIs.

2 Materials and methods

2.1 Study area

Our study sites are situated along the western-central basin of Lake Erie, the fourth largest of the five Great Lakes in North America and the eleventh largest lake globally (Hansen 1989). Lake Erie is located on the international boundary between the United States and Canada. The northern shore is bounded by the Ontario province of Canada, while the US states of Michigan, Ohio, Pennsylvania, and New York bound the western, southern, and eastern shores. The three study sites, Crane Creek (CRC), Portage River (PTR), and Old Woman Creek (OWC) (Fig. 1), are located in the North West Ohio portion of the Western Lake Erie Basin (WLEB), which is one of United States’ most significant collections of inland rivers and streams. The WLEB covers nearly 7 million acres and stretches across most of northwest Ohio, portions of northeast Indiana, and southeast Michigan. Around 75% of the land is used for agricultural production. Approximately 1.2 million people live in the basin, distributed between three urban centers, Toledo, Ohio; Fort Wayne, Indiana; and Lima, Ohio, and numerous cities and towns.

Fig. 1
figure 1

Map of the United States (top left), surficial geologic map of the study area showing the dominant soil parent material overlaid on hill shade base map (right). Data source: Gridded Soil Survey Geographic Database for Ohio (SSURGO 2012), with dominant soil parent material overlay by Subburayalu and Slater (2013)

Full size image

The geology of the Lake Erie region is characterized by middle Paleozoic sedimentary rocks composed of limestones, dolomites, shales, and sandstones (Bolsenga and Herdendorf 1993). These rocks were deposited about 430 to 300 million years ago under conditions ranging from tropical barrier reef habitats to deltaic and deepwater clastic environments associated with mountain-building (orogenic) episodes and tectonic plate collisions (Herdendorf 2013). The age of the bedrock units in this coastal region ranges from the Silurian Period (416 to 435 million years ago) in western Ohio to the Pennsylvanian Period (307 to 318 million years ago) in the bedrock highland areas (see Appendix: Fig. 13) (ODNR 2018). Along the Lake Erie shore west of Sandusky, bedrock units exposed at the surface or buried beneath glacial deposits are mostly Silurian and Devonian-age limestone and dolomite (exposed at Catawba, Bass, and Kelleys Islands). East of Sandusky, Devonian-age shale trends along the shore into northeastern Ohio (exposed in the valley walls of the Vermilion, Black, and Rocky Rivers).

Myers et al. (2000) and ODNR (2018) classified the soils across the three study sites into Lakebed (lacustrine) soils and glacial till soils based on their parent materials. The lakebed soils are fine-textured lacustrine deposits usually formed at the lake bottoms and were deposited during the prehistoric stages of Lake Erie’s formation (ODNR 2018). Glacial till soils are unsorted (variable-sized) materials that were mixed, crushed, compressed, and transported by the movement of glaciers. Till soils have variable textures and can be slightly permeable below the surface. Also, these soils can be classified further into Inceptisols, Alfisols, Mollisols, and a small fraction of Entisols based on the dominant soil order (see Appendix: Fig. 14).

2.2 Lithostratigraphy

Two-inch diameter piezometers were installed at the upland, transition, and wetland zones of each of the three field sites, using a hand auger with soil samples retrieved every 0.1 m. The piezometers were deeper at the upland zones terminating at about 6 m, while the transition and wetland piezometers terminated at about 2 m and 1 m, respectively. The soil samples retrieved during piezometer installation were used to create lithostratigraphic logs (lithologs) that were used to ground-truth some of the geophysical measurements.

2.3 Electromagnetic induction

The EMI method measures the response of the ground to the propagation of electromagnetic fields made up of an alternating electric intensity and magnetizing force. An alternating current is passed through a transmitter coil (a loop of wire) placed over the ground to generate a primary (inducing) magnetic field which spreads out both above and below the ground surface. In a homogeneous ground, the primary field is detected by a receiver coil with a minor reduction in amplitude (Halder 2018). In the presence of a conducting body, however, the magnetic component of an electromagnetic field penetrating the ground induces the flow of eddy currents within the conductor. The eddy currents generate their own secondary electromagnetic field, which differs in phase, amplitude, and direction, sensed by the receiver coil. The differences between transmitted and received electromagnetic fields reveal the presence of a conductor and provide information on its geometry and electrical properties (Geonics Limited 2016; Gebbers et al. 2009). For this study, EMI data was acquired with an EM38-MK2 sensor (Geonics, Canada). The sensor consists of a transmitter and two receiver coils at separation distances of 0.5 m and 1.0 m from the transmitter and outputs apparent electrical conductivity (ECa) at average depth ranges of 0–0.75 m and 0–1.5 m in vertical mode and 0–0.38 m and 0–0.75 m in horizontal mode. However, the true penetration depth of the sensor depends on the sensor frequency and conductivity of the topsoil, which is site-specific (e.g., Paton and PENSERV Corp Pa. 2012). The sensor operates at a frequency of 14.5 kHz and delivers ECa values in mS/m.

In this study, the EM38-MK2 sensor was used in vertical mode to pace around each site with a back-mounted real-time kinematic differential ground positioning system (RTK-DGPS). The RTK-DPS system was set up using two Emlid Reach RS2+ differential GPS (Emlid Ltd., Hong Kong), one was fixed at a location that serves as the base while the other was mounted on a backpack and serves as the rover; this allowed the acquisition of a georeferenced data at about 0.3 m accuracy. The data acquisition was monitored real-time using EM38-MK2win data logging system operated on a Windows 10 based field tablet computer. The EMI system was nulled and calibrated at each site before data acquisition, and the sensor was held up at about 0.4 m from the ground during acquisition. TrackMaker38MK2 (Geonics, Canada) was used to process the raw data and georeference it with the GPS data, the outputted xyz data file was imported into Surfer 12 (Golden Software, Colorado, USA), to enable the creation of grid files and interpolation of the data using inverse distance to a power approach (Franke 1982), yielding a map of spatially distributed ECa. At the CRC transition and wetland zones, measurements were repeated in December 2022 and April 2023 to investigate the temporal variability of the soil ECa.

2.4 Electrical resistivity tomography

Electrical resistivity tomography is used to determine the subsurface distribution of electrical resistivity by carrying out a set of resistance measurements on the ground surface and/or in boreholes. Current is injected into the ground via two current electrodes, and the resulting potential difference is measured at another two electrodes using different combinations of current and potential electrodes along a transect or grid. A geophysical inversion of the acquired data is then performed to obtain the resistivity of the subsurface (see Loke 1997). In this study, ERT data were collected across the three sites with a SuperSting R8 resistivity meter and an 84-electrode switch box (Advanced Geosciences Inc., Austin, TX), using the dipole–dipole electrode configuration (e.g., Loke 1997) and 1 m unit electrode spacing. The data was collected in automatic mode, which automatically records resistivity data using a preprogrammed command file and the distributed Swift automatic multi-electrode system (AGIUSA 2006). At the CRC site, ERT data were collected along 3 transects in the upland zone and another 3 transects between the transition and wetland zones. At the PTR site, six different transects were used to acquire ERT data. The two longest profiles were acquired using a roll-along method up to a total spread of 147 m from the transition zone to the wetland zone and 168 m from the upland zone to the wetland zone, while the other four transects were 84 m long. The ERT data at the OWC site were acquired along 7 different transects cutting across the three zones; thus, a total of 19 resistivity profiles were obtained across the three sites. The ERT survey was designed in such a way as to enable a correlation of electrical resistivity with the lithostratigraphic logs obtained from the piezometers installed in each of the sites. The processing (inversion) of the acquired resistivity data was performed with the AGI EarthImager 2D (Advanced Geosciences Inc., Austin, TX) using a smoothness constrained inversion method. The resulting resistivity profiles were used to build fence diagrams using RockWorks software (RockWare Inc, Colorado), in order to improve the spatial visualization of the ERT results. Finally, the Earth Imager was used to trim some of the ERT profiles to a depth suitable for high-resolution correlation to be made with the well logs.

2.5 Ground penetrating radar

Ground penetrating radar is a geophysical method that uses propagating electromagnetic waves to investigate the shallow subsurface based on its response to changes in the electromagnetic properties of the shallow subsurface. The propagation wave velocity is determined by the relative permittivity contrast between different soil layers or the background material and anomalous body (e.g., Baker et al. 2007). The transmitter component of the GPR system propagates the electromagnetic wave through the earth material and the interactions with the earth material response are sensed by the receiver component. The GPR survey in this study was carried out on short survey lines, collocated on some of the ERT survey lines in each site to allow the comparison of both methods in terms of suitability for investigating vertical variations and delineating subsurface heterogeneity at the land-lake interface. GPR data were collected using PulseEKKO GPR system (Sensors & Software Inc., Canada) with a 200 MHz antenna. Transmitter and receiver separation of 0.5 m was used, and the GPR data was collected at 0.5 m intervals using a manual trigger method. The new DVL-500 ruggedized display unit (a high-visibility touchscreen) was used to visualize the data simultaneously during acquisition.

The acquired data were processed using Sensor & Software’s EKKO_Project, following standard GPR processing for subsurface characterization (e.g., Annan 2009), to remove low-frequency noise due to inductive coupling effects and/or dynamic range limitations of the antennas (Annan 2009).

2.6 Soil and groundwater measurements

Teros 12 soil sensors, which measure soil moisture, temperature, and electrical conductivity (Meter Group, Inc., USA), were installed at 10 and 30 cm depth in the upland (n = 4 and 2, respectively), transition (n = 4 and 2, respectively), and wetland (n = 2 and 2, respectively) zones of each site and were used to monitor monthly soil moisture (SM) changes between March 2022 and April 2023.

Also, each piezometer was instrumented with Aqua TROLL 600 multiparameter sondes (In-situ Inc., USA), which were used to measure monthly changes in groundwater level and specific conductivity. The sensors were equipped with wipers to minimize fouling of sensor heads and calibrated according to manufacturer protocols during maintenance visits. The soil and groundwater sensors were both set to log data on a 15-min frequency.

3 Results

3.1 Spatial variability of soil properties from apparent electrical conductivity

The sites show high spatial variability in the ECa distribution from the 0.5 m and 1.0 m sensor separation (Table 1), corresponding to average depths of 0–0.75 m and 0–1.5 m, respectively (Fig. 2). Also, the ECa values showed Gaussian distribution across all sites, with higher values in the wetland and transition zones than the upland zones for the PTR and OWC sites. The CRC upland showed higher ECa values at 0.5 m (mean = 55.3; variance = 0.041) and 1.0 m (mean = 49.3; variance = 0.012) sensor separation than the CRC transition and wetland at 0.5 m (mean = 12.8; variance = 0.004) and 1.0 m (mean = 28.7; variance = 0.006) sensor separation. The ECa values across the sites are generally higher at the 1.0 m coil separation than at 0.5 m (see Table 1), but again the CRC upland showed an opposite behavior with higher ECa values at the 0.5 m sensor separation (Fig. 2A, Table 1). The CRC transition and wetland zones also showed lower ECa values compared to OWC and PTR wetland and transition zone.

Fig. 2
figure 2

Apparent electrical conductivity (ECa) distribution maps of the sites A CRC upland, B CRC wetland and transition, C PTR, and D OWC, at transmitter–receiver spacing of 0.5 m (top) and 1.0 m (bottom) which correspond to approximate depth of 0–0.75 m and 0–1.5 m, respectively

Full size image
Table 1 The ECa distribution across the study sites at 0.5 m and 1.0 m coil spacing
Full size table

3.2 Soil ECa patterns compared to traditional soil maps

Previous works have recommended that ECa maps be used to optimize soil mapping (e.g., Corwin and Lesch 2003; Mertens et al. 2008). Soil ECa maps are compared to traditional soil maps from the United States Department of Agriculture (USDA), as shown in Fig. 3. A closer match between the USDA soil maps and the soil ECa maps was observed at CRC upland and OWC sites than for PTR and CRC wetland and transition. Generally, the ECa maps revealed soil units that were identified from the USDA soil maps, and also revealed the presence of minor subunits that were not captured in the traditional soil maps (Fig. 3). At the CRC site, Toledo silty clay (To), Toledo silty clay, ponded (Tp), and Nappanese silty clay loam (NpA) were the three major soil units identified from the USDA soil map (Fig. 3A), both To and Tp are hydric soils with 0–1% slope while NpA is a non-hydric soil with 0–3% slope. The hydric soil units Tp and To showed higher ECa values than the non-hydric NpA soil unit. The ECa maps provided a more precise detail of the lateral extension of each of these soil units than the soil map and identified additional subunits that were missing in the soil map (see Appendix Table 2). At the CRC wetland and transition, the USDA soil map placed the site in one soil unit (Tp), while the ECa map showed a clearer lateral variation indicating the presence of additional units/sub-units (Fig. 3B). The PTR soil map showed only the To and Tp soil units (Fig. 3C), with the ECa values higher in Tp than the To, as was also the case at CRC upland site. The ECa map at the PTR site showed a more precise lateral extent of the Tp and To soil units and also indicated the presence of additional units/sub-units which were missed out in the soil map (see Appendix: Table 2). At OWC, the soil map showed three distinct soil units which were clearly identified by the ECa maps (Fig. 3D), the Zurich silt loam (ZuF), which is rated non-hydric with 25–40% slope, Holly silt loam (HoA) which is rated hydric with 0–1% slope (USDA-NRCS 2018), and Del Rey silt loam (DeA), a nearly level and somewhat poorly drained soil with 0–2% slope. The hydric HoA soil unit showed higher ECa values than the non-hydric units.

Fig. 3
figure 3

An overlay of soil ECa distribution from 1.0 m spaced sensors on USDA soil maps. A, B CRC upland and CRC transition and wetland showing three soil units: Toledo silty clay (To), Toledo silty clay, ponded (Tp), and Nappanese silty loam (NpA). C PTR showing two soil units: Toledo silty clay (To) and Toledo silty clay, ponded. D OWC showing three soil units, Zurich silt loam (ZuF), Holly silt loam (HoA), and Del Rey silt loam (DeA). Data source: USDA soil maps obtained from Web Soil Survey (see USDA 2019)

Full size image

3.3 Soil moisture and groundwater dynamics

In situ soil moisture, temperature, and electrical conductivity (EC) data were compared across the upland, transition, and wetland zones of the CRC site (Fig. 4). At the wetland zone, the SM increased slightly from 0.52 to 0.55 m3 m−3 at 10 cm depth and from 0.45 to 0.46 m3 m−3 at 30 cm depth between December 2022 and April 2023. At the transition zone, the SM also increased between December 2022 and April 2023, and the values ranged from 0.26 to 0.32 and 0.46 to 0.50 m3 m−3 at 10 cm depth and from 0.30 to 0.40 and 0.41 to 0.44 m3 m−3 at 30 cm depth (Fig. 4C). The changes in soil moisture were much more variable at the top 10 cm. During the same period, we recorded a substantial increase in soil electrical conductivity from about 10 to 750 uS/cm and 500 to 1000 uS/cm in the transition and wetland zones, respectively (Fig. 4A); the soil temperature was also close to 0 °C as shown in Fig. 4B.

Fig. 4
figure 4

A Soil electrical conductivity (EC), B temperature, and C moisture changes recorded between April 2022 and April 2023 at CRC upland zone (UP), transition zone (TR), and wetland zone (W)

Full size image

In August 2022, the specific conductivity of groundwater at the CRC transition zone decreased with the hydraulic head (Fig. 5A). The water level in the piezometers continued to decrease until it dried up in October–December 2022 (no data recorded). In the same period, our results show a decrease in both soil moisture (Fig. 4C) and soil electrical conductivity (Fig. 4A), while the soil temperature increased from April and peaked in August before decreasing to a minimum around December for both transition and wetland zones (Fig. 4B). In April 2023, the specific conductivity showed a steady increase with the hydraulic head (Fig. 5B).

Fig. 5
figure 5

Specific conductivity and hydraulic head variations recorded in A August 2022 and B April 2023, at the CRC transition zone

Full size image

3.4 Vertical variability of soil properties assessed from ERT and GPR

The ERT results show a vertical variation in electrical resistivity along 19 different transects across the three sites, generally increasing from the soil surface to 19.7 m (Fig. 6). The CRC site showed resistivity values that ranged from 5.1 to 54.4 Ωm and 10.4 to 70 Ωm for the upland and wetland-transition zones, respectively (Fig. 6A, B). A low resistivity layer is clearly visible at the depth of 1.3–6 m in the upland and 0–3 m in the transition and wetland zones. At the PTR site, the resistivity ranged from 5.5 to 74 Ωm (Fig. 6D), with low resistivity values from the soil surface to a depth of about 6 m in the transition zone and about 3 m in the wetland zone. In the upland zone, higher resistivity values are observed from 0 to 1.6 m depth. The OWC showed a more variable resistivity response, which ranged from 10.2 to 148 Ωm (Fig. 6C); high resistivity values were observed in the upland but also in the transition and wetland at shallower depths compared to the CRC and PTR sites. While the CRC and PTR sites are generally flat, the OWC site shows significant elevation differences between the upland zone and the wetland or transition zone. The ERT profiles 2 and 3, which extended from the wetland into the upland, were corrected for terrain effect during inversion.

Fig. 6
figure 6

Electrical resistivity tomography profiles from the sites. A CRC upland, B CRC transition and wetland, C OWC site, and D PTR site. For clarity, the profiles in each zone are numbered as P1, P2…Pn, according to the sequence of acquisition. The various ERT transects results are arranged in a fence diagram to reveal the lateral and vertical variation in subsurface electrical resistivity

Full size image

The GPR reflection radargram revealed three distinct stratigraphic layers at the CRC transition (Appendix: Fig. 10A). Layer 3 showed stronger reflection compared to layers 1 and 2. Similarly, three stratigraphic layers were also identified at the PTR site, on a transect which extended from the upland zone to the wetland zone (Appendix: Fig. 10B). There is a visible lateral change in GPR reflection at about 45 m mark along the profile, which signifies the boundary between upland soil (1) and wetland soil (2) as described in Fig. 10B (see Appendix). There are also some vertical features that appeared at 10 m and 60–70 m along the profile and extended to an estimated depth of about 3–3.5 m, which are probably a strong reflection of till, as the soil samples retrieved from piezometer installation in this zone confirmed that the till here is very rich in pebbles (diamictites) composed mainly of black shale. At the OWC site, two distinctive layers were observed. A top layer (1) of about 0.5 m thickness which showed a weak reflection and a second layer (2) with a stronger reflection which lies between 0.5 and 1.5 m (Appendix: Fig. 10C). The GPR data is useful to identify the stratigraphic boundaries at these sites but does not reveal what the structures are. Combining different geophysical methods is useful to overcome this challenge by leveraging the strength of each method to bridge the gap in interpretation where other methods are lacking. Thus, the GPR results will be compared with other methods to better identify the observed layers and structures (see Section 3.6).

3.5 Lithostratigraphy reconstructed from well logs

The lithostratigraphy of the three sites is described here based on the lithologs obtained from the piezometers in upland, transition, and wetland zones of each site (Fig. 7). The upland zones are characterized by silty clay layers at the top 1.7 m, underlain by two distinct layers of clay with intercalations of black shale and claystone (glacial till) which extended down to 5.4 m at CRC and 5.5 m at PTR sites (see Fig. 7: BH1 and BH4). At the OWC site, the glacial till layers extended to 3.3 m depth underlain by 3.7 m thick rhythmite clay (Fig. 7: BH7). The CRC transition zone shows similar stratigraphy with the CRC upland, while the PTR transition zone is different from the upland, it shows a layer of silty clay which extends down to 0.5 m followed by a clay layer down to 2 m. The OWC upland zone is characterized by a thin layer of silty loam at the top 6 cm, followed by a clay layer extending down to 1.35 m, then a silty clay from 1.35 to 4.0 m, followed by water-saturated clay from 4 to 5.8 m. The wetland zones of the three sites show similar stratigraphy characterized by a 1 m thick clay layer.

Fig. 7
figure 7

The lithostratigraphy of the sites described based on soil cores retrieved from the piezometers from the upland zones (BH1, BH4, and BH7), the transition zones (BH2, BH5, BH8), and wetland zones (BH3, BH6, and BH10) of the CRC, PTR, and OWC sites, respectively, with depth in meters. BH9 is from OWC wetland edge, with soil properties identical to that of the OWC transition

Full size image

3.6 Combining ERT, lithostratigraphic logs, and GPR

At the CRC upland zone, the lithologs reconstructed from piezometer BH1 (see Fig. 7) were correlated with two ERT profiles that cut across the piezometer at 34.5 m (Fig. 8A) and at 64 m (Fig. 8B). The stratigraphic boundaries observed in the lithologs matched that of the ERT profiles. The BH1 log identified a sharp boundary in the till layer marked by a different clay-to-rock fragment ratio; this boundary was also observed in the ERT profiles (Fig. 8A, B). The low resistivity layer in the upland is tied to the till, while the higher resistivity layer is tied to the silty clay layer. In the transition and wetland zones, the stratigraphic boundaries observed in the lithologs (BH2 and BH3) also matched that of the ERT profiles taken across them (Fig. 8C–E). Figure 8C showed a transect from the wetland zone to the transition zone, the lithologs and ERT results identified a shift from clay-dominated top layer in the wetland to silty clay-dominated top layer in the transition zone. The till layer appeared deeper in the wetland zone (Fig. 8E) compared to parts of the transition zone (Fig. 8D).

Fig. 8
figure 8

ERT profiles correlated with lithologs from piezometers at the Crane Creek (CRC) site, showing a close match between stratigraphic boundaries from lithologs and that of ERT. A, B The dark blue and light blue layers indicate two distinct layers of till identified at CRC upland, the dark blue till layer is clay-rich while the light blue is gravel-rich. C–E The till layer at CRC transition zone is composed of more gravel than clay and thus showed higher resistivity than the surrounding wet clay. The piezometer at CRC wetland is not deep enough to get into the till

Full size image

At the PTR site, the litholog from the upland piezometer (BH4) was tied to two ERT profiles (see Appendix: Fig. 11A and D). The stratigraphic boundaries observed from the well log also matched that of the ERT profiles very closely; the high resistivity layer observed from the ERT profile was linked to a layer of dry silty clay. The litholog from the transition zone piezometer (BH5) was tied to an ERT profile that ran from the upland zone into the transition zone (Appendix: Fig. 11B) and another that ran from the transition zone into the wetland zone (Appendix: Fig. 11E). The top layer of dry silty clay identified from the litholog clearly matched the ERT result. In the wetland zone, both the ERT profile and the litholog (BH6) identified a top layer of wet clay (Appendix: Fig. 11C). Since the wetland piezometer is just 1 m deep, it was not possible to determine the thickness of this clay layer from the litholog alone, but correlating it with the ERT profile showed the thickness to be between 2.5 and 3.2 m in the wetland zone.

At the OWC site, the upland zone showed high resistivity layer at the top 1.6 m, which correlated with dry silty clay based on the litholog from the upland piezometer (BH7); this was underlain by two layers of till which is consistent with CRC and PTR upland zones. The litholog also revealed a continuous layer of rhythmite clay from 3.3 to 6 m. The transition zone showed relatively uniform resistivity at the top 3 m which is tied to a silty clay layer based on the litholog obtained from the piezometer (BH8) at 2 m depth (see Appendix: Fig. 12D). The wetland edge also showed a less heterogeneous layer at the top 3 m which is tied to silty clay as well based on piezometer (BH9) data (Fig. 12A and C). The wetland center showed lower resistivity response compared to the transition zone; this low resistivity unit was found to be a wet clay layer when tied to the piezometer (BH10) data obtained at 1 m depth (Fig. 12A and B). These results indicate that the stratigraphy and soil moisture dynamics are the key drivers of spatial heterogeneities at these sites.

A comparison between collocated ERT and GPR profiles showed that GPR also provided information about vertical variability of soil properties at the study sites. However, the GPR sensitivity at these sites is limited to the top 1.5 m, probably due to signal attenuation (conductive losses) resulting from the high clay content at these sites, while the ERT clearly showed better depth resolution. Correlating the GPR results (Appendix: Fig. 10B) and ERT results (Appendix Fig. 11D) was useful to confirm that the vertical features observed in the GPR just below the dry silty clay are glacial till (Fig. 9B); this clearly shows that combining different geophysical methods is a better approach for subsurface characterization than using one method alone.

Fig. 9
figure 9

Comparing collocated ERT and GPR profiles at A CRC transition, B PTR upland to wetland transect, and C OWC wetland edge to wetland center transect. Combining the GPR and ERT methods helped to clearly identify the clay, silty loam, silty clay, and till layers

Full size image

4 Discussion

4.1 Spatiotemporal variation of soil properties across land-lake interfaces

In this study, we hypothesized that ECa will be low in the upland zones and increase as we move from the upland to the transition zone and from the transition to the wetland zone. The EMI results from PTR and OWC agreed with our hypothesis (Fig. 2C and D), but higher conductivity was observed at the CRC upland (Fig. 2A) compared to its transition and wetland (Fig. 2B). Comparing the EMI results at CRC site with the monthly soil moisture data measured at the site revealed that soil moisture is the key driver of lateral variation in ECa.

Previous studies have found that the ECa correlates strongly with soil moisture (SM) and organic matter (OM) (e.g., Molin and Faulin 2013; Shanahan et al. 2015). A recent study by Emmanuel et al. (2023) found that SM correlated with OM, which further suggests that both parameters are somewhat interdependent and thus challenging to decouple. Although some studies observed a slightly stronger influence of OM on soil ECa (e.g., Shanahan et al. 2015; Emmanuel et al. 2023), Domsch and Giebel (2004) argued that SM has a stronger influence on soil ECa, which is probably the case in some wetlands considering that Emmanuel et al. (2023) found their strongest correlation with silt content, which correlated better with SM (r2 = 0.660) than with OM (0.632). Measured soil ECa is known to be driven by both static and dynamic factors which includes soil moisture content, bulk density, soil salinity, clay content and mineralogy, and temperature (Corwin and Lesch 2005). Also, Johnson et al. (2005) noted that the ECa magnitude and spatial variability at any site is usually dominated by one or more of these factors, and also vary from site to site. It follows that sites where dynamic factors (e.g., soil moisture and salinity) dominate, temporal changes in spatial patterns usually show less stability than sites dominated by static factors (e.g., soil texture and silt content). In static factor dominated sites, variation in soil dynamic properties affects only the ECa magnitude, while the spatial pattern remains consistent (Johnson et al. 2005). In our study, we fail to observe a consistent spatial pattern (see Appendix: Fig. 13) which further indicates that the measured ECa at our sites are dominated by dynamic factors; this agrees also with the results from the soil sensors.

At our sites, it is possible that an increase in SM due to groundwater level rise could have led to changes in soil water chemistry (e.g., dilution), which would have resulted in the observed temporal variation of the ECa at the CRC transition and wetland zones (see Fig. 15). SM varies spatially both laterally across a site and vertically through the soil profile; this is well described by Corwin and Lesch (2005). At these sites, SM was mostly higher at the top 10 cm than at 30 cm depths across the sites except for the periods between June and December when SM was higher at 30 cm (Fig. 4C). This implies that the variation of ECa values between 0.5 and 1.0 m sensor separation is due to variation in SM. EMI could therefore serve as a non-invasive tool for monitoring soil water dynamics to understand ground water-soil water exchanges and their control on biogeochemical processes at land-lake interfaces.

The close match between the USDA soil maps and the soil ECa maps observed at CRC upland and OWC sites indicates that the EMI is useful for soil mapping and can be relied upon at sampling restricted sites. Furthermore, the ECa maps reveal that all the hydric soils across the sites (To, Tp, and HoA) have high ECa values as shown in Fig. 3 and Table 2 (see Appendix); this implies that the EMI could help soil scientists and ecologist to non-invasively map the lateral extent of hydric soils.

4.2 Geophysics can help reconstruct subsurface stratigraphy of land-lake interfaces

To test the suitability of geophysical methods for characterizing subsurface stratigraphy of land-lake interfaces, we investigated the sites using ERT, GPR, and lithostratigraphic logs (lithologs) from piezometers. The ERT results from the three sites (Fig. 6) showed relatively lower resistivity values at CRC and PTR in the range of 5.1–74 Ωm than at the OWC site (10.2–148 Ωm). Though the existing wells at the OWC wetland were not deep enough to ground-truth the high resistivity values observed in the wetland zone, existing borehole data close to the upland area revealed the presence of dry till at the depth of 50 ft (15 m) that could explain the observed higher resistivity. The correlation of the ERT data with lithologs was very useful in understanding what drives the vertical variation in electrical resistivity across the sites. For example, the high resistivities observed close to the surface in the upland zones of PTR (Appendix: Fig. 11) and OWC (Appendix: Fig. 12) sites were linked to dry silty clay. The results also indicate that similar soil types could show different resistive responses at different sites depending on how their electrical property compares to that of the surrounding material. For example, a layer of till showed very low resistivity values at the CRC upland, while at the PTR site the same till layer appeared as a relatively more resistive layer. This is because, at the CRC upland, the till is surrounded by more resistive silty clay layers, while at the PTR site, the till is overlaid by wet clay layers, which is much less resistive than the till layer.

The GPR method also revealed structural heterogeneity at the study sites; it clearly identified the boundaries between silty loam and clay layers at CRC and OWC and between dry silty clay and clay at the PTR sites. However, the GPR sensitivity at these sites is limited to the top 1.5 m compared to ERT, which provided high-resolution depth sensitivity up to 19.7 m using a 1 m electrode spacing. The poor resolution observed with GPR above 1.5 m depth could be linked to signal attenuation due to high clay content. SM is known to cause signal attenuation in GPR data (e.g., Huisman et al. 2003; Klotzsche et al. 2018; Agbona et al. 2021). This signal attenuation due to water saturation is expected to be more pronounced in wetlands with high water residence time and could also lead to temporal variations in GPR measurement due to seasonal variation in SM in such wetlands.

The surficial geology map of the study area shown in Fig. 1 indicates that the geology of the area is characterized by lakebed soils (fine textured lacustrine deposits), underlain by glacial till soils. This agrees with the ERT results of this study, which revealed that the stratigraphy is made up of silty clay and clay layers which are lake bed soils (fine textured lacustrine deposits) and two different types of till layers, differing in their composition (Fig. 7), which the surficial geology map revealed to be clayey Wisconsin till and loamy Wisconsin till. These tills are rich in clay, which are expected to slow down infiltration and, thus, increase water residence time, which could help to sustain diverse biogeochemical exchanges at these land-lake interfaces. The electrical response of these TAI soils depends on whether they are dry or saturated; this explains why silty clay layers are very resistive in upland areas where they are very dry (see Appendix: Fig. 11), and also why glacial till showed lower resistivity at the CRC site (Fig. 8A and B) and high resistivity at PTR site (Appendix: Fig. 11). These results indicate that geophysical methods are useful to reconstruct subsurface stratigraphy of land-lake interfaces.

4.3 The pros and cons of combining different geophysical methods to study land-lake interfaces

No single geophysical method is capable of capturing all the complexities of soil state and processes particularly in a dynamic TAI ecosystem. In this study, our expectation is that combining the EMI, ERT, and GPR will help to overcome the limitations associated with each method. For example, the EMI is very useful to map lateral variation in soil properties but the depth resolution is limited to about 1.5 m, while ERT and GPR have an advantage of higher depth resolution along a transect but lacks high spatial resolution, in addition to the signal attenuation associated with GPR in wet soils. With only the EMI, we would have missed the vital information about vertical variability in soil properties which helped us to delineate the stratigraphy of the land-lake interface. Although our results show the advantage of combining multiple methods, one key challenge to consider is the time required to complete the study with all the methods. Since the land-lake interface is a dynamic zone with various processes occurring at different time scales, it is very important that the measurement with different methods be conducted within the same time frame to minimize the effect of temporality that might bias the comparison.

5 Conclusions

This work demonstrates that despite the unique properties of the land-lake interface and its dynamic processes, geophysical methods are useful to understand its soil architecture and subsurface stratigraphic heterogeneities at a higher spatial resolution than point sampling methods. The close match between ECa maps and USDA soil maps, as well as the additional details provided by the ECa maps, implies that EMI is a useful tool for optimizing soil mapping and could also be used to extrapolate soil properties, particularly at sampling-restricted sites where only non-invasive measurements are feasible.

Unlike the aggregate and the pore approaches of investigating soil architecture which focuses on studying limited core samples, the geophysical methods show a more detailed characterization of soil spatial heterogeneity, with good lateral and vertical resolution. The EMI provided better lateral heterogeneity at high resolution, while ERT and GPR provided high-resolution vertical variation in the soil profile. The stratigraphy of these land-lake interfaces and their soil moisture dynamics were found to be the key drivers of the observed heterogeneities. Future studies should consider a detailed investigation of temporal variability of the geophysical signals coupled with monitoring temporal changes in SM and soil water quality to better understand the mechanism behind the temporal variation of the ECa observed here and quantify the influence of fluctuating SM and groundwater levels on the geophysical measurements.