Wetlands are unique ecosystem units that are seasonally or permanently inundated (Davidson et al., 2018). Their high-water residence time and diverse microbiomes allow them to perform important hydrological and biogeochemical functions including nutrient retention and transformation (Ameli and Creed, 2017; Mitsch and Gosselink, 2000). Hence, wetlands are used to mitigate nutrient loading into surface water bodies to prevent eutrophication (Fisher and Acreman, 2004; Griffiths and Mitsch, 2020). In Northwest Ohio, areas which were originally part of the historic Great Black Swamp, a large wetland drained in the mid- nineteenth century to establish human settlement and farmlands are being restored back into wetland conditions in efforts to mitigate nutrient loading into Lake Erie (Lenhart and Lenhart, 2014; Luckeydoo et al., 2002; Mitsch and Day, 2006). The hydrological and biogeochemical functioning of restored wetlands depend to a large extent on their soil types and properties which show high spatial and temporal variation (Richardson and Craft, 2020; Walton et al., 2020). Besides a high level of heterogeneity generally associated with soils (Chatterjee et al., 2021; Quine and Zhang, 2002), the spatial and seasonal variation in soil saturation and flooding within wetlands create variations in aerobic and anaerobic conditions in wetlands soils (Jackson et al., 2014; Reddy et al., 2013). Also, as reservoirs retaining water and solutes eroded and transported from upland, wetland soils have high and varying carbon and nutrient accumulation rates (Craft et al., 2018; Mitsch et al., 2013). These non-stationary processes within wetlands drive their high heterogeneity with both spatial and temporal variation in their soils’ biological and physicochemical properties. Wetland soils serve as long-term integrators of nutrient loading effects (Reddy et al., 2013; Reilly et al., 2021), hence, investigating their properties distribution including texture, bulk density, porosity, moisture content and organic matter concentration is a vital step in the pre- and post- restoration characterization and monitoring of wetlands.
Classically, field estimation of soil properties involves collecting soil samples either in a disturbed form (e.g., using augers) or minimally disturbed as soil cores and performing the appropriate laboratory tests to obtain the property of interest (Pennock et al., 2008; Reddy et al., 2013; Sabbe and Marx, 1987). There has also been an increasing advancement in the use of in situ technologies to measure soil properties of interest such as the use of Time Domain Reflectometry (TDR) and Near-Infrared Spectrometry (NIRS) to measure soil moisture and organic matter contents respectively (Gehl and Rice, 2007; Walker et al., 2004; Wenjun et al., 2014). While these methods give direct estimates of the soil property of interest, they are point based and would require a high number of samples to characterize a large area at high spatial and temporal resolution (Doro et al., 2013). Collecting and analyzing high numbers of soil samples or installing a large array of in situ sensors is not practical due to the prohibitive cost, soil disturbance, and field effort needed. Also, intensively collecting soil samples at wetlands is particularly challenging due to their complex terrain, seasonal inundation, and the need to minimize disturbances at these sites. Hence, wetland soil property estimation will benefit from the use of proximal and remote sensing techniques using geophysical and or remote sensing methods (Casa et al., 2013; Doro et al., 2021; Ge et al., 2011; Romero-Ruiz et al., 2018).
Geophysical techniques such as electrical resistivity, ground penetrating radar and electromagnetic imaging have been widely used to characterize the distribution of soil properties at field scale with applications to agriculture (Turkeltaub et al., 2022; Vingiani et al., 2022), geotechnical engineering (Xie et al., 2022), and contaminated sites management (Atekwana and Atekwana, 2010; Boudreault et al., 2010; Cassiani et al., 2014). Among these techniques, electromagnetic imaging which is used to obtain the spatial distribution of soil apparent electrical conductivity (ECa) and apparent magnetic susceptibility (MSa) has been more popular due to available instruments that do not require direct contact with the subsurface and can be towed behind farm vehicles (Corwin and Lesch, 2005a; b; Doolittle and Brevik, 2014; Heil and Schmidhalter, 2017). Measurements of soil ECa correlates with soil texture, soil moisture, organic matter, salinity, cation exchange capacity and nutrient concentration (Kweon et al., 2013; Martínez et al., 2010; Martini et al., 2017; Molin and Faulin, 2013; Rhoades et al., 1990). Using these correlation relationships, the spatial distribution of soil ECa has been used to predict the distribution of soil properties (Kelley et al., 2017; Stępień et al., 2015), characterize variation in soil crop yield (Corwin and Scudiero, 2020), and to monitor changes in soil moisture in response to precipitation and extended evapotranspiration (Martínez et al., 2010; Martini et al., 2017; Shanahan et al., 2015). Besides the statistical correlation between ECa and soil properties (Huang et al., 2017; Wang et al., 2022), petrophysical relationships have also been developed relating electrical conductivity with soil properties (Chan et al., 2000; Lesmes and Friedman, 2005; Tso et al., 2019; Weller et al., 2013). In this regards, the Archie’s law (Archie, 1942) and several of its modified forms which relate measured bulk electrical conductivity to the electrolytic and surface conduction of current in soils have been used to estimate soil moisture distribution and changes in salinity with time (Comas and Slater, 2004; Binley and Slater, 2020; Glover, 2010). While the direct use of measured soil ECa distribution is common, geophysical inversion codes including EM4Soil (Bonomi et al., 2001; Triantafilis and Monteiro Santos, 2013) and EMagPy (McLachlan et al., 2021a) for estimating the true electrical conductivity (EC) distribution from measured ECa are available. Solving the inverse problem allow for estimating the distribution of soil EC with depth given different simultaneous ECa measurement with different transmitter and receiver spacing which give different investigation depths. While this is desirable, this study only uses the measured ECa and MSa. The inversion of measured ECa to estimate vertical distribution of electrical conductivity and assess their correlations with the vertical distribution of soil properties is planned in a subsequent study.
Although EMI using instruments such as the Geonics EM-38-MK2 can output both ECa and MSa, most soil properties studies using EMI ignores the measured MSa as limited correlation has been observed between the MSa and soil properties (Heil et al., 2017; Sadatcharam et al., 2020). This is related to the limited penetration depth and the difficulty with interpreting the in-phase component of the EMI signal which is related to the MSa (De Smedt et al., 2014). There is however sufficient evidence in the literature showing variation in soil magnetic susceptibility (MS) and suggesting the dependence of bulk MS on soil properties including particle size, organic matter content, cation exchange capacity as well as biogeochemical translocation processes (César de Mello et al., 2020; de Jong et al., 2000; Lecoanet et al., 1999; Marques et al., 2014; Ramos et al., 2017). Grimley and Vepraskas (2000) presented an application of the variation in MS for delineating wetland hydric soils. They showed that bulk MS values were relatively lower in hydric soils compared to non-hydric soils due to the seasonal saturation, high organic matter content, anerobic and Fe-reducing environment typical of hydric soils. While the study by Grimley and Vepraskas (2000) were based on laboratory analysis of soil samples using a Bartington magnetic susceptibility meter, model MS2, and an MSF probe (Bartington Instruments Ltd., Witney, Oxford, UK), field applications of MS for delineating wetlands with MS meters have been documented (Grimley et al., 2004; Grimley et al., 2008; Simms and Lobred, 2011). These studies validate the capabilities of using MS to study wetland soil properties distributions. However, no study has been done to assess the use of MSa data measured along ECa during EMI to obtain additional information on wetland soil properties.
The wide use of electromagnetic imaging (EMI) for soil characterization has mainly focused on agricultural and geotechnical engineering applications (Corwin and Lesch, 2005b) with limited applications to wetlands (McLachlan et al., 2021b). There is increasing interest in restoring and constructing wetlands globally and especially within the Great Lakes region because of the desire to protect the lakes from eutrophication due primarily to phosphorus runoff. Extending the application of EMI to wetland soil characterization could provide wetland scientists and managers with a non-invasive tool for delineating varying wetlands soil properties prior to restoration and construction and during post restoration monitoring. With the documented extensive literature showing the use of measured ECa for characterizing soil nutrients (e.g., NO3−) (Heiniger et al., 2003; Korsaeth, 2005), adapting EMI to wetlands could provide a non-invasive, albeit indirect means for assessing nutrients (NO3− and PO4−) retained in wetlands soil. This is a major goal for several of the wetland restoration initiatives in the Midwestern United States including the H2Ohio initiative in Ohio, USA (Berkowitz et al., 2021; Jurjonas et al., 2022; Mitsch and Wang, 2000).
In this study, we present an application of EMI for characterizing wetlands’ soil properties distribution at a field scale. We use direct estimates of soil properties including soil texture (clay, silt, and sand contents), bulk density, porosity, moisture content, and organic matter concentration estimated from laboratory analysis of limited soil core samples to assess their statistical correlation with measured ECa and MSa. We compared the ECa and MSa distribution to a USDA soil map of the area to assess how well these data captures known soil variation within the wetland. Using least square linear regression models between ECa and the soil properties validated using a leave-one-out cross validation technique, we use soil ECa distribution to predict the SMC and SOM distribution at the study site. With wetlands showing high and varying organic matter contents (Ahn and Jones, 2013; Bai et al., 2005; Bishel-Machung et al., 1996; Bruland and Richardson, 2006; Saeed et al., 2019), we particularly assessed the influence of soil organic matter on soil apparent electrical conductivity which has received limited attention from the literature.