The study site is part of the Lehstenbach catchment located in the Fichtelgebirge mountains (Fig. 1a), a low mountain range in northeastern Bavaria (Germany) close to the border with the Czech Republic. Various soil types including Dystric Cambisols, Haplic Podsols, and Histosols (i.e., peat soil) cover the catchment area of approximately 4.2 km2, situated on top of variscan granite bedrock (Strohmeier et al., 2013). The catchment is bowl shaped (Fig. 1b), and minerotrophic riparian peatlands have developed around the major streams. The plot-scale study site (Fig. 1c) is located in a riparian peatland draining into a nearby stream close to the catchment's outlet (Fig. 1b).

The groundwater level in this area annually varies within the top 30 cm of the peat soil, and the local groundwater flow has a S–SW orientation (Durejka et al., 2019) towards a nearby drainage ditch. Permanently high water saturation of the peat soil favors the development of anoxic biogeochemical processes close to the surface. Frei et al. (2012) suggested that hotspots in the study area are related to the stimulation of iron-reducing bacteria and accumulation of iron sulfides, which are generated by small-scale subsurface flow processes and the spatially non-uniform availability of electron acceptors and donors induced by the typical micro-topography of the peatland. Non-uniform availability of electron acceptors and donors in combination with labile carbon stocks is the primary driver in generating biogeochemical hotspots in the peatland (Frei et al., 2012; Mishra and Riley, 2015).

The experimental plot for the geophysical measurements covers approximately 160 m2 (12.6 × 12.6 m, Fig. 1c) of the riparian peatland. Sphagnum Sp. (peat moss) and Molinia caerulea (purple moor grass) dominate the vegetation, with the sphagnum and purple moor-grass abundance being higher in the northern part of the plot (Fig. 2a and b). In the southeastern region, where the sphagnum is less abundant, permanent surface runoff was observed (Fig. 2d). Peat thickness was measured with a 1 m resolution in the E–W direction and 0.5 m resolution in the N–S direction (along the IP profiles described below). To measure the thickness of the peat, a stainless-steel rod (0.5 cm in diameter) was pushed into the ground until it reached the granitic bedrock (similarly to Parry et al., 2014). The local groundwater level was measured in two piezometers and was found at ∼ 5–8 cm below the surface during the IP survey. Groundwater samples were collected at three different locations (S1, S2, and S3 indicated in Fig. 3) using a bailer. Porewater profiles were taken at S1, S2, and S3 at 5 cm intervals to a maximum depth of 50 cm below ground surface (b.g.s.) using stainless-steel mini-piezometers. All water samples were filtered through 0.45 µm filters and analyzed for fluoride, chloride, nitrite, bromide, nitrate, phosphate, and sulfate using an ion chromatograph (Compact IC plus 882, Metrohm GMBH). Dissolved organic carbon (DOC) was measured using a Shimadsu TOC analyzer via thermal combustion. Dissolved iron species (Fe2+) and total iron (Fetot) concentrations were measured photometrically using the 1,10-phenanthroline method on porewater samples that had been stabilized with 1 % vol / vol 1 M HCl in the field (Tamura et al., 1974). Two freeze cores (see Fig. 2d) were extracted at locations S1 and S2 (Fig. 3) by pushing an 80 cm long stainless-steel tube into the peat. After the tube was installed, it was filled with a mixture of dry ice and ethanol. After around 20 min, the pipe with the frozen peat sample was extracted and stored on dry ice for transportation to the laboratory at the University of Bayreuth. Both freeze cores were cut into 10 cm segments. Each segment was analyzed for reactive iron (1 M HCl extraction and measured for Fetot as described above) (Canfield, 1989), reduced sulfur species using the total reduced inorganic sulfur (TRIS) method (Canfield et al., 1986) and carbon and nitrogen concentrations after combustion using a thermal conductivity detector. Peat samples were also analyzed by FTIR using a Vector 22 FTIR spectrometer (Bruker, Germany) in absorption mode with subsequent baseline subtraction on KBr pellets (200 mg dried KBr and 2 mg sample). Thirty-two measurements were recorded per sample and averaged from 4500 to 600 cm-1 in a similar manner to Biester et al. (2014).

The induced polarization (IP) imaging method, also known as complex conductivity imaging or electrical impedance tomography, is an extension of the electrical resistivity tomography (ERT) method (e.g., Kemna et al., 2012). As such, it is based on four-electrode measurements, where one pair of electrodes is used to inject a current (current dipole) and a second pair of electrodes is used to measure the resulting electrical potential (potential dipole). Modern devices can measure tens of potential dipoles simultaneously for a given current dipole, permitting the collection of dense data sets within a reasonable measuring time. This provides an imaging framework to gain information about lateral and vertical changes in the electrical properties of the subsurface. IP data can be collected in the frequency domain (FD), where an alternating current is injected into the ground where the polarization of the ground leads to a measurable phase shift between the injected periodic current and the measured voltage signals. From the ratio of the magnitudes of the measured voltage and the injected current as well as the phase shift between the two signals, we can obtain the electrical transfer impedance. The inversion of imaging data sets, i.e., a large set of such four-point transfer-impedance measurements collected at different locations and with different spacing between electrodes along a profile, permits us to solve for the spatial distribution of the electrical properties in the subsurface (see deGroot-Hedlin and Constable, 1990; Kemna et al., 2000; Binley and Kemna, 2005).

IP inversion results can be expressed in terms of the complex conductivity (σ*) or its inverse the complex resistivity (σ*=1/ρ*). The complex conductivity can be denoted either in terms of its real (σ′) and imaginary (σ′′) components or in terms of its magnitude (|σ|) and phase (ϕ): 1σ*=σ′+iσ′′=|σ|eiϕ, where i=-1 is the imaginary unit, σ=σ′2+σ′′2, and ϕ=tan⁡-1(σ′/σ′′). The real part of the complex conductivity is mainly related to the Ohmic conduction, while the imaginary part is mainly related to the polarization of the subsurface materials. The conductivity (σ′) is related to porosity, saturation, the conductivity of the fluid filling the pores, and a contribution of the surface conductivity (Lesmes and Frye, 2001). The polarization (σ′′) is only related to the surface conductivity taking place at the electrical double layer (EDL) at the grain–fluid interface. For a detailed description of the IP method, the reader is referred to the work of Ward (1988), Binley and Kemna (2005), and Binley and Slater (2020).

The strongest polarization response is observed in the presence of electrically conducting minerals (e.g., iron) (e.g., Pelton et al., 1978) in the so-called electrode polarization (Wong et al., 1979). It arises from the different charge transport mechanisms in the electrical conductor (electronic or semiconductor conductivity) and the electrolytic conductivity of the surrounding pore fluid, which make the solid–liquid interface polarizable. Diffusion-controlled charging and relaxation processes inside the grain (e.g., Revil et al., 2018, 2019; Abdulsamad et al., 2019) or outside the grain in the electrolyte (e.g., Wong, 1979; Bücker et al., 2019) are considered possible causes of the polarization response at low frequencies irrespective of the specific modeling approach. All mechanistic models predict an increase in the polarization response with increasing volume content of the conductive minerals (Wong, 1979; Revil et al., 2015a, b, 2017a, b, 2018; Qi et al., 2018; Bücker et al., 2018).

In the absence of electron conductors, the polarization response is only related to the accumulation and polarization of ions in the EDL. Different models have been proposed to describe the polarization response as a function of grain size, surface area, and surface charge (e.g., Schwarz, 1962; Schurr, 1964; Leroy et al., 2008). Alternatively, the membrane polarization related the IP response to variations in the geometry of the pores as well as the concentration and mobility of the ions (e.g., Marshall and Madden, 1959; Bücker and Hördt, 2013; Bücker et al., 2019). Regardless of the specific modeling approach, EDL polarization mechanisms strongly depend on the specific surface area of the material and the charge density at the surface (Revil, 2012; Waxman and Smits, 1968).

In this study, we conducted FDIP measurements at 1 Hz along 65 lines during a period of 4 d in July 2019. We used the DAS-1 instrument manufactured by Multi-Phase Technologies (now MTP-IRIS Inc.). We collected 64 N–S-oriented lines (referred to as By 1 to By 64) with 20 cm spacing between each line. One additional line (By 68) was collected with a W–E orientation, which intersects the N–S-oriented lines at 3 m, as presented in Fig. 3. Each profile consisted of 64 stainless-steel electrodes (3 mm diameter) with a separation of 20 cm between each electrode (Fig. 2c). Besides the short electrode spacing, the use of a dipole–dipole configuration with a unit dipole length of 20 cm warranted a high resolution within the upper 50 cm of the peat, where the biogeochemical hotspots were expected. We deployed a dipole–dipole skip-0 (i.e., the dipole length for each measurement is equal to the unit spacing of 20 cm) configuration; voltage measurements were collected across eight adjacent potential dipoles for each current dipole. The dipole–dipole configuration avoids the use of electrodes for potential measurements previously used for current injections to avoid contamination of the data caused by remnant polarization of electrodes. To evaluate data quality, reciprocal readings were collected along one profile every day (see, e.g., LaBrecque et al., 1996; Flores Orozco et al., 2012a, 2019). Reciprocal readings refer to data collected after interchanging current and potential dipoles. We used coaxial cables to connect the electrodes with the measuring device to minimize the distortion of the data due to electromagnetic coupling and cross-talking between the cables (e.g., Zimmermann et al., 2008, 2019; Flores Orozco et al., 2013), with the shields of all coaxial cables running together into one ground electrode (for further details, see Flores Orozco et al., 2021).

The principle of reciprocity asserts that normal and reciprocal readings should be the same (e.g., Slater et al., 2000). Hence, we use here the analysis of the discrepancy between normal and reciprocal readings to detect outliers and to quantify data error (LaBrecque et al., 1996; Flores Orozco et al., 2012a; Slater and Binley, 2006; Slater et al., 2000). In this study, we quantified the error parameter for each line collected as normal and reciprocal pairs (using the approach outlined by Flores Orozco et al., 2012a) and computed the average value of the error parameters for the different lines to define the error model used for the inversion of all imaging data sets.

For the inversion of the IP imaging data set, we used CRTomo, a smoothness-constrained least-squares algorithm by Kemna (2000) that allows inversion of the data to a level of confidence specified by an error model. We used the resistance and phase error models described by Kemna (2000) and Flores Orozco et al. (2012a). The resistance (R) error model is expressed as s(R)=a+bR, where a is the absolute error, which dominates at small resistances (i.e., R<0), and b is the relative error, which dominates at high resistance values (LaBrecque et al., 1996; Slater et al., 2000). For the phase, the error model is also expressed as a function of the resistance s(ϕa)=cRd, where d<0 in our study due to the relatively low range in the measured resistances (see Flores Orozco et al., 2012a, for further details). If d→0, the model reduces to the constant-phase-error model (Flores Orozco et al., 2012a) with s(ϕ)=c described by Kemna (2000) and Slater and Binley (2006).

In Fig. 4, we present a modified pseudosection showing both normal (negative levels) and reciprocal (positive levels) readings in terms of apparent resistivity (ρa) and apparent phase (ϕa) for the data collected along line By 25. Plots in Fig. 4 show consistency between the normal and reciprocal readings of apparent resistivity (a) and phase (b). Figure 4c and d show the histograms of the normal–reciprocal misfits along line By 25 for both the resistance and phase (ΔR and Δϕ, respectively), which exhibit near-Gaussian distributions with low standard deviations (as expected for random noise) for both the normalized resistance (SR=0.027) and the apparent phase (Sϕ=1.1 mrad). Readings exceeding these standard deviation values were considered outliers (between 16 % and 33 % of the data at the different lines) and were removed from the data set prior to the inversion.

Here, we present inversion results obtained using the error parameters, a=0.001Ω, b=0.022, and c=1 mrad. For the imaging, we defined a cut-off value of the cumulated sensitivity of 102.75, with pixels related to a lower cumulated sensitivity blanked in the images. The cumulated sensitivity values are a widely used parameter to assess the depth of investigation (Kemna et al., 2002; Flores Orozco et al., 2013).

The thickness of the peat in the plot was found to vary between 40 and 160 cm (Fig. 5). The thickness of the peat unit increased in the W–E direction, with much smaller variations in the N–S direction. Variations in the vegetation cover (as indicated by the three vegetation classes, abundant av, moderate mv, and sparse sv) do not seem to correspond to the variations in the peat thickness. Note that the N–S orientation of the majority of IP lines is approximately aligned with the direction of minimum changes in the peat thickness.

Figure 6 shows the imaging results of the N–S-oriented profiles By 25 and By 46 and the W–E-oriented profile By 68 expressed in terms of the conductivity (σ′) and polarization (σ′′). These images reveal three main electrical units: (i) a shallow peat unit with high σ′ (> 5 mSm-1) and high σ′′ (> 100 µS m-1) values in the top 10–20 cm b.g.s., (ii) an intermediate unit in the peat with moderate to low σ′ (< 5 mS m-1) and moderate σ′′ (40– 100 µS m-1) values, and, (iii) underneath it, a third unit characterized by moderate to low σ′ (< 5 mS m-1) and the lowest σ′′ (< 40 µS m-1) values corresponding to the granite bedrock. The compact structure of the granite, corresponding to low porosity, explains the observed low conductivity values (σ′ < 5 mS m-1) due to low surface charge and surface area. The shallow and intermediate electrical units are related to the relatively heterogeneous peat (Fig. 6), which is beyond the vertical change and lateral heterogeneities in the complex conductivity parameters. As shown in the plots of σ′′ in Fig. 6, the contact between the second and third units roughly corresponds to the contact between peat and granite measured with the metal rod. This indicates the ability of IP imaging to resolve the geometry of the peat unit. However, for the survey design used in this study, σ′′ images are not sensitive to materials deeper than ∼ 1.25 m. Images of the electrical conductivity reveal much more considerable variability and a lack of clear contrasts between the peat and the granite materials, likely due to the weathering of the shallow granite unit (Lischeid et al., 2002; Partington et al., 2013).

The phase of the complex conductivity represents the ratio of the polarization relative to the Ohmic conduction (ϕ=σ′′/σ′). Thus, it can also be used to represent the polarization response (Kemna et al., 2004; Ulrich and Slater, 2004; Flores Orozco et al., 2020). Similarly to the σ′′ images, the phase images presented in Fig. 7 resolve the three main units: (i) the shallow peat unit within the top 10–50 cm is characterized by the highest values (ϕ > 18 mrad), (ii) the intermediate unit still corresponding to peat is characterized by moderate ϕ values (between 13 and 18 mrad), and (iii) the third unit, associated with the granitic bedrock, is related to the lowest ϕ values (< 13 mrad). The polarization images expressed in terms of ϕ show a higher contrast between the peat and the granite units than the σ′ (or σ′′) images. The histograms presented in Fig. 7 show the distribution of the phase values in the images, with a different color for model parameters extracted above and below the contact between peat and granite. The histograms highlight the fact that the lowest phase values clearly correspond to the granite bedrock (< 13 mrad), while higher phase values are characteristic of the peat unit.

Moreover, the shallow unit shows more pronounced lateral variations in the phase than in σ′′, and patterns within the peat unit are more clearly defined. As observed in Fig. 6, along line By 25, the thickness of the first unit decreases from approx. 0.5 m at 2 m along the profile to 0 m around 10 m at the end of the profile. Along line By 46, the first unit is slightly thicker than 50 cm and shows the highest phase values (∼ 25 mrad) between 0 and 6.5 m along the profile. Beyond 6.5 m, the polarizable unit becomes discontinuous with isolated polarizable (∼ 18 mrad) zones, extending to a depth of 50 cm. The geometry of the shallow, polarizable unit is consistent with the corresponding results along line By 68, which crosses By 25 and By 46 at 3 m along these lines (S1 and S3 are located at these intersections). In particular, the highest phase values are consistently found in the shallowest 50 cm in the peat unit, at the depth where biogeochemical hotspots have been reported in the study by Frei et al. (2012).

Figure 8 presents maps of the electrical parameters at different depths aiming to identify lateral changes in the possible hotspots across the entire experimental plot. Such maps present the interpolation of values inverted in each profile. Along each profile, a value is obtained through the average of model parameters (conductivity magnitude and phase) within the surface and a depth of 20 cm (shallow maps) and between 100 and 120 cm (for deep maps). The western part of the experimental plot (between 0 and 4 m in the x direction and between 2 and 9 m in the y direction) corresponds to a shallow depth to the bedrock (a peat thickness of ∼ 50–70 cm) and is associated with high electrical parameters in the shallow maps (ϕ > 18 mrad, σ′ > 7, and σ′′ > 100 mS m-1), which we can interpret here as the geometry of the biogeochemical hotspots. Another hotspot can be identified in the northern part of the experimental plot, in the area with abundant vegetation; we observe a higher polarization response for the top 20 cm (ϕ > 18 mrad and σ′′ > 80 µS m-1) than, for instance, the one corresponding to the moderate vegetation located in the southern part. In contrast, the lowest polarization values (ϕ < 15 mrad and σ′′ < 80 µS m-1), which we interpret as biogeochemically inactive zones, are related to the area with sparse vegetation and permanent surface runoff.

Kleinebecker et al. (2009) suggest that besides climatic variables, biogeochemical characteristics of the peat influence the composition of vegetation in wetlands. Hence, we can use variations in the vegetation as a qualitative way of evaluating our interpretation of the IP imaging results. In Fig. 8g–i, we present the histograms of the electrical parameters extracted at each of the three vegetation features defined in the experimental plot (abundant, moderate, and sparse). These histograms show, in general, that the location with sparse vegetation, i.e., with permanent surface runoff, is related to the lowest phase values (histogram peak at 13 mrad). Moderate vegetation corresponds to moderate phase and σ′′ values (histogram peak at 18 mrad and 70 µS m-1, respectively). In comparison, the abundant vegetation corresponds to the highest phase and σ′′ values (histogram peak at 22 mrad and 90 µS m-1, respectively) in the top 20 cm. The histogram of the three vegetation features in terms of σ′ values overlaps with each other.

The evaluations of the imaging results measured along profiles By 25, By 46, and By 68 were used to select the locations for the freeze core and sampling of groundwater. Sampling points S1 and S3 were defined in the highly polarizable parts of the uppermost peat unit (high σ′ and σ′′ values). In contrast, sampling point S2 is located in an area characterized by low polarization values. Figure 9a–e show the chemical parameters measured in the water samples, specifically chloride (Cl-), phosphate (PO43-), dissolved organic carbon (DOC), total iron (Fetot= Fe2++ Fe3+), and pH, whereas Fig. 9f–j show the chemical parameters measured in the peat samples extracted from the freeze cores, namely, CEC, concentrations of iron sulfide (FeS or FeS2), total reactive iron (Fetot), potassium (K+), and sodium (Na+). The pore-fluid conductivity measured in water samples retrieved from the piezometers shows minor variation with values ranging between 6.7 and 10.4 mS m-1. To facilitate the comparison of electrical parameters and geochemical data, Fig. 9k–m show the complex conductivity parameters (σ′, σ′′, and ϕ) at the sampling points S1, S2, and S3, which were extracted as vertical 1D profiles from the corresponding imaging results.

As observed in Figs. 6 and 7, the highest complex conductivity values (σ′, σ′′) were resolved within the uppermost 10–20 cm and rapidly decreased with depth. Furthermore, the values of ϕ and σ′′ in the top 20 cm at S1 and S3 are significantly higher than those at location S2. High values of ϕ and σ′′ at S1 and S3 correspond to high concentrations of DOC, phosphate, Fetot in water samples, as well as high K+, and Na+ contents measured in soil materials extracted from the freeze cores. Figure 9 reveals consistent patterns between geochemical and geophysical parameters: in the first 10 cm b.g.s. close to sampling points S1 and S3, we observe complex conductivity values (σ′ and σ′′) as well as chemical parameters, such as DOC and phosphate (only at S1). Accordingly, at S1 Fetot also reveals at least 2 times higher concentrations than those measured in S2.

Figure 10 shows the actual correlations between the complex conductivity and Cl-, DOC, and Fetot concentrations measured in groundwater samples. In Fig. 10, we also provide a linear regression analysis to quantify the correlation between parameters. Figure 10 reveals that the phase has a weak to moderate correlation with DOC, Cl-, and Fetot. The conductivity (σ′) shows a slightly stronger correlation with the DOC, the Cl-, and total iron concentration than the polarization (σ′′). The highest σ′′ values (> 100 µS m-1) are related to the highest DOC and total iron concentration.

Further evidence of the presence of the biogeochemical hotspot interpreted at the position of S1 is available by the FTIR spectroscopy analysis of the freeze-core samples presented in Fig. 11. The spectra show the absorbance intensity at different wave numbers, C–O bond (∼ 1050 cm-1), C=O double bond (∼ 1640 cm-1), carboxyl (∼ 1720 cm-1), and O–H bonds (∼ 3400 cm-1). The peaks are also indicated in Fig. 11 with the interpretation based on the typical values reported in peatlands, for instance, McAnallen et al. (2018) or Artz et al. (2008).

The geochemical and geoelectrical parameters presented in Figs. 6–7 and 9 reveal consistent patterns, with the highest values within the uppermost 10 cm around S1 and S3. The high DOC, K+, and phosphate concentrations in the uppermost peat layers and especially in the areas found to be biogeochemically active strongly suggest that there is rapid decomposition of dead plant material in these areas (Bragazza et al., 2009). Ions such as K+ and phosphate are essential plant nutrients, and phosphate species especially are often the primary limiting nutrient in peatlands (Hayati and Proctor, 1991). The presence of dissolved phosphate in porewaters suggests that (i) the plant uptake rate of this essential nutrient is exceeded by its production through the decomposition of plant material and (ii) that organic matter turnover must be rapid indeed to deliver this amount of phosphate to the porewater. This is supported by the DOC concentrations in porewater exceeding 10 mM. DOC is produced as a decomposition product during microbial hydrolysis and oxidation of solid-phase organic carbon via enzymes such as phenol oxidase (Kang et al., 2018). Enzymatic oxidation processes are enhanced by oxygen ingress via diffusion and, more importantly, by water table fluctuations that work as an “oxygen pump” to the shallow subsurface (Estop-Aragonés et al., 2012). Thus, an increased DOC concentration in the porewater can be used as an indicator of microbial activity (Elifantz et al., 2011; Liu, 2013). The small amount of phosphate measured in the less active area S2 can be explained by advective transport from the active area S1, which is directly “upstream” of S2. In this case, advective water flow through the uppermost peat layers along the hydrological head gradient may have transported a small amount of reaction products from the biogeochemical source areas to the “non-active” area. The high DOC, Fe, K+, and phosphate (only at S1) levels confirm our initial interpretation of the highly conductive and polarizable geophysical units within the first 20–50 cm b.g.s. in the surroundings of S1 and S3 as biogeochemically active areas.

The high DOC concentrations are also likely to be directly or indirectly responsible for the Fe maximum in the upper layers. Dissolved Fe was predominantly found as Fe2+ (reducing conditions), suggesting either that high labile DOC levels maintain a low redox potential or that the dissolved Fe2+ was complexed with the DOC limiting the oxidation kinetics enough so that Fe2+ can accumulate in peat porewaters. The TRIS analysis clearly showed very low levels of sulfide minerals in both freeze cores, especially in the uppermost peat layers. This was unexpected considering the reducing conditions implied by the dominance of porewater Fe2+. We argue that the lack of sulfide minerals is due to insufficient H2S or HS- needed to form FeS or FeS2 or that the redox potential was not low enough to reduce sulfate to H2S or HS-. Both mechanisms are possible, as groundwater in the catchment generally has low sulfate concentrations, and yet sulfate was detected in peat porewater samples, which would not be expected if redox potentials were low enough to reduce sulfate to sulfide. The chemical analyses do not reveal any significant or systemic vertical gradient in mineral sulfide concentrations, as expected for the site (Frei et al., 2012). The maximum in extractable (reactive) solid-phase Fe was also located in the uppermost peat layer at the “hotspot” S1. This Fe was likely in the form of iron oxides or bound to/in the plant organic matter. Such iron-rich layers typically form at the redox boundary between oxic and anoxic zones and can be highly dynamic depending on variations in the peatland water levels and oxygen ingress (Wang et al., 2017; Estop-Aragonés et al., 2013).

Similarly to other peatlands (Artz et al., 2008), the FTIR spectra show the presence of carbon–oxygen bonds such as C–O, C=O, and COOH at both S1 and S2. Furthermore, the peak intensities at S1 tend to decrease with the depth, while the peak intensities at S2 samples tend to increase in agreement with the increase in the polarization response (both phase and σ′′). This observation further supports our interpretation of the shallow 10 cm in IP images in the vicinity of S1 as a biogeochemical hotspot. However, such a biogeochemical hotspot is not related to the accumulation of iron sulfides, which was suggested by Abdel Aal and Atekwana (2014) or Wainwright et al. (2016) as the main parameter controlling the high IP response. The phosphate and Fe could potentially form complexes with the O–H groups that show an absorbance peak at 1050 cm-1 (Arai and Sparks, 2001; Parikh and Chorover, 2006). Furthermore, the iron can also form complexes with the carboxyl groups (absorbance at ∼ 1720 cm-1).

The two electrical units observed within the peat indicate variations in the biogeochemical activity with depth. Thus, it is likely that the anomalies associated with the highest σ′, σ′′, and ϕ values in the uppermost unit correspond to the location of active biogeochemical zones, i.e., a hotspot. Consequently, the moderate σ′, σ′′, and ϕ values indicate a less biogeochemically active or even inactive zone in the peat. The third unit represents the granitic bedrock. The low metal content and the well-crystallized form of the granite lead to low σ′′ values (here, < 40 µS m-1), as suggested by Marshall and Madden (1959).

The high polarization response of the biogeochemically active peat (here σ′′ > 100 µS m-1 and ϕ > 18 mrad) is consistent with the measurements of McAnallen et al. (2018), who performed time-domain IP measurements in different peatlands. They suggest that the active peat is less polarizable due to the presence of the abundant sphagnum cover. They found that in the areas where the peat is actively accumulating, the ratio of the vascular plants and the non-vascular sphagnum is low and, therefore, the oxygen availability is low. However, the sphagnum is expected to exude a small amount of carbon into the peat, and Fenner et al. (2004) found that the sphagnum contributes to the DOC leachate to the porewater, which is contradictory to the model of McAnallen et al. (2018). In agreement with Fenner et al. (2004), in our study, we also observe that high DOC content correlates with abundant sphagnum cover, which is also found in conjunction with abundant purple moor grass. In this regard, recent studies have demonstrated an increase in the polarization response due to the accumulation of biomass and activity in the root system (e.g., Weigand and Kemna 2017; Tsukanov and Schwartz, 2020). However, the sphagnum does not have roots; thus, it cannot directly contribute to the polarization response. McAnallen et al. (2018) suggest that the vascular purple moor grass can contribute to the high IP, as the roots transport oxygen into the deeper area, increasing the wettability and normalized chargeability of the peat.

Derived from the results and discussion above, we delineated the geometry of the hotspots. The map presented in Fig. 12 is based on the maps of phase and imaginary conductivity values at depths of 10 and 20 cm. Hotspots interpreted in those areas exceeded both a phase value of 18 mrad and imaginary conductivity of 100 µS m-1 at the same time. Besides the geometry of the hotspots, Fig. 12 indicates that the hotspot activity attenuates with the depth.

In this study, we have found a strong correlation between the polarization response (ϕ and σ′′) and Fetot in the solid phase and a less pronounced correlation between the polarization response and the concentration of dissolved iron in the liquid phase (see Fig. 10). In all considered mechanistic polarization models, the phase value depends on the volumetric content of metallic particles (Wong, 1979; Revil 2015a, b; Bücker et al., 2018, 2019; Feng et al., 2020) and, therefore, the phase could reveal the possible metallic content in the peat. If the iron in the solid phase occurred in the form of highly conductive minerals, the two above correlations would point to the polarization mechanism of perfect conductors described by Wong (1979) as a possible explanation for the observed response. Previous studies (e.g., Flores Orozco et al., 2011, 2013) attributed the polarization of iron sulfides (FeS or FeS2) in sediments to such a polarization mechanism as long as sufficient Fe2+ cations are available in the porewater. Such an effect has been investigated in detail by Bücker et al. (2018, 2019) regarding the changes in the polarization response due to surface charge and reaction currents carried by redox reactions of metal ions at the mineral surface. However, in the case of the present study, the lack of sulfide and the rather high pH (inferred from the presence of sulfate) in the porewater do not favor the precipitation of conductive sulfides such as pyrite. Under these conditions, iron would rather precipitate as iron oxide or form iron–organic matter complexes. The electrical conductivity of most iron oxides is orders of magnitude smaller than the conductivity of sulfides (e.g., Cornell and Schwertmann, 1996) and is thus too low to explain an increased polarization based on a perfect-conductor polarization model (e.g., Wong, 1979; Bücker et al., 2018, 2019; Feng et al., 2020). The only highly conductive iron oxide is magnetite, with a conductivity similar to pyrite (Atekwana et al., 2016). Consequently, the presence of magnetite could explain such a polarization. However, the low pH (∼ 5) typical for peat systems does not favor the precipitation of magnetite, but rather less conducting iron (oxy)hydroxides such as ferrihydrite (Andrade et al., 2010; Linke and Gislason, 2018). Analysis of sediments of the freeze core also did not reveal magnetite. As indicated by the FTIR analysis, the iron might furthermore have built complexes with the carboxyl (absorbance at ∼ 1720 cm-1). Such moderately conductive iron minerals or iron–organic complexes might still cause a relatively strong polarization response as predicted by the polarization model developed by Revil et al. (2015) and Misra et al. (2016a). In this model, which attributes the polarization response to a diffuse intra-grain relaxation mechanism, the polarization magnitude is mainly controlled by the volumetric content. In this model, the (moderate) particle conductivity only plays a secondary role (e.g., Misra et al., 2016b).

The product of both surface charge density and specific surface area can be quantified by the CEC of a material. As peat mainly consists of organic matter known to have a high CEC, even when compared to most clay minerals (e.g., Schwartz and Furman, 2014, and references therein), the polarization of charged organic surfaces may explain the observed IP response. Additionally, Garcia-Artigas et al. (2020) concluded that bioclogging due to fine particles and biofilms increases the specific surface area and the CEC, resulting in an increase in the polarization response. However, the CEC values measured in samples retrieved from the freeze core vary in a narrow range between 5 and 25 meq kg-1, and we did not observe any correlation between CEC and changes in the polarization magnitude (σ′′, ϕ). Such a lack of correlation between the polarization effect and the CEC was also reported by Ponziani et al. (2011), who conducted spectral IP measurements on a set of peat samples. Hence, the measured CEC is high enough to explain a rise in EDL polarization; however, the (small) variation in CEC does not explain the observed variation in the polarization magnitude.

The pH of the pore fluid is also known to control the magnitude of EDL polarization; an increase in pH usually corresponds to an increase in the polarization magnitude (e.g., Skold et al., 2011). At low pH values, H+ ions occupy (negative) surface sites and thus reduce the net surface charge of the EDL (e.g., Hördt et al., 2016, and references therein). Our data seem to show the opposite behavior: we found a lower pH in the highly polarizable anomalies at S1 and S3 compared to site S2 (the inactive and less polarizable location), while the pH increases at depth for decreasing values in the polarization (both σ′′ and ϕ). At the same time, variations in pH are within the range 4.45 and 5.77 and thus might not be sufficiently large to control the observed changes in the polarization response.

Besides pH, pore-fluid salinity plays a significant role in the control of EDL polarization. Laboratory measurements on sand and sandstone samples indicated that an increase in salinity leads to an early increase in the imaginary conductivity, which is eventually followed by a peak and a decrease at very high salinities during later stages of the experiments (e.g., Revil and Skold, 2011; Weller et al., 2015). Hördt et al. (2016) provided a possible theoretical explanation for this behavior: in their membrane-polarization model, salinity controls the thickness of the diffuse layer of the EDL and depends on the specific geometry of the pores; there is an optimum thickness, which maximizes the magnitude of the polarization response. In the present study, we observed that an increase in salinity (as indicated by the high Cl- concentrations within the uppermost 10 cm at all sampling locations) is associated with an increase in the polarization magnitude response (e.g., Revil and Skold, 2011; Weller et al., 2015; Hördt et al., 2016). However, the highest Cl- concentrations were observed for the shallow layers at location S2, where we measured lower polarization magnitudes (in terms of σ′′, ϕ) compared to S1 and S3.

The strong correlation between the polarization response and the DOC suggests an, as yet not fully understood, causal relationship. A similar observation has recently been reported by Flores Orozco et al. (2020), who found a strong correlation between the organic carbon content as a proxy of microbial activity and both σ′ and σ′′ in a municipal waste landfill in Austria. Regarding the available carbon, McAnallen et al. (2018) reported a strong correlation between the occurrence of long-chained C=O double bonds and the total chargeability of peat material. The upper peat layers are exposed to oxygen, leading to oxidation of the peat and formation of C=O double bonds at solid-phase surfaces and in the porewater DOC. Such long-chained organic molecules have an increased wettability and thus more readily attach (or even form at organic matter surfaces) to the surface of solid organic and mineral particles (Alonso et al., 2009). Based on a membrane-polarization model, Bücker et al. (2017) predict an increase in the polarization magnitude in the presence of wetting (i.e., long-chained) hydrocarbon in the free phase. The long-chained polar DOC attaches to the peat surface, similarly to polar hydrocarbon, and so it might provide extra surface charge, thus reducing the pore space and causing membrane polarization (Marshall and Madden, 1959).

As suggested by Vindedahl et al. (2016), organic matter can adsorb to the iron-oxide surface via electrostatic attraction and provides a negatively charged macromolecular layer on the iron oxide. Such complexes could also explain the observed increase in the polarization response in the anomalies interpreted as biogeochemical hotspots. The point of zero charge of the peat is below pH 4 (Bakatula et al., 2018), while for iron (oxide) it varies between ∼ 5 and ∼ 9 (Kosmulski et al., 2003). This means that the organic matter is probably negatively charged, and the iron oxide is most likely positively charged since the measured pH at the sample points varies between 4.5 and 5.8, with lower values in the top 10 cm in the hotspot area. Hence, in the shallow 10 cm from S1 and S3, the pH favors the DOC to bond with the iron in the solid phase.