Integrated use of georadar, electrical resistivity, and SPT for site characterization and water content estimative

Abstract Geophysical methods are potent tools for geotechnical site characterization in a non-destructive way. They improve the extrapolation of punctual data from direct survey methods, allowing a fast and cost-effective evaluation of large areas. Ground Penetrating Radar (GPR) and DC electrical resistivity (ER) are the most requested methods for geotechnical and geoenvironmental applications. Their use, however, is usually uncoupled, with no sharing of information from one method to another to improve data interpretation. This case study illustrates the development of protocols and scripts in R© programming language for ER and GPR data analysis with Standard Penetration Tests (SPT) data to produce more accurate information on subsurface conditions concerning lithology, water content, and groundwater table (GWT) position. The SPT data were used to associate resistivity ranges with different soil lithologies and GPR pulse velocities for estimating the soil water content. Estimated water content values aided in interpreting ER data and locating the groundwater table. The contacts between layers in the radargrams allowed the refinement of the ER model, rendering 3D volumes for each soil layer in situ.


Introduction
Geotechnical and geoenvironmental applications use the Ground Penetrating Radar (GPR) and electrical resistivity geophysical investigation methods intensively.They provide the possibility of extrapolating punctual geotechnical surveys, allowing a better and less expensive subsurface characterization (Cosenza et al., 2006).GPR is a non-destructive subsurface investigation method based on the propagation and reflection of electromagnetic pulses.It has proven to be a high-resolution, time and cost-efficient subsurface imaging method for geotechnical applications.GPR, however, presents limitations in the case of high-conductivity soils because of their elevated electrical conductivity, which causes GPR's signal attenuation.GPR's most common geotechnical applications are related to the subsurface location of buried structures such as pipes, galleries, and tanks and detection of different soil layer interfaces (Souza & Gandolfo, 2012).The reflections caused by an object crossing the radargram plane will form a hyperbola (Davis & Annan, 1989), a function of the velocity of propagation of the electromagnetic pulse (V), as illustrated in Equation 1: where x is the horizontal distance between the antenna's center and the object's center, z is its depth, and t is the two-way travel time.The value of V is a function of the electromagnetic properties of the propagation medium, and it can be estimated using an expression derived from Maxwell's equation (Equation 2): where C is the light propagation in vacuum (3.10 8 m/s), σ is the medium electrical conductivity (S/m), ω is the angular frequency (rad/s), μ (μ = 4π x 10 -7 T m/A in vacuum, μ o ) is the magnetic permeability, and ε is the electric permittivity (ε = 8.854 C 2 /N m 2 in a vacuum, ε o ) (Topp & Davis, 1981;Topp et al., 1980 The contacts between layers in the radargrams allowed the refinement of the ER model, rendering 3D volumes for each soil layer in situ. frequencies used by GPR (0.02 to 2.5 GHz), the medium conductivity has a minor influence on the pulse propagation velocity, which can be approximated by Equation 3 (Davis & Annan, 1989): where ε r = ε/ε o is the relative dielectric constant of the medium.Table 1 illustrates typical values of relative dielectric constant, electrical conductivity, and propagation velocity for different materials.As observed, water content significantly influences the pulse propagation velocity because of its high dielectric constant (ε r = 80).
The interval velocity between two reflectors (V int ) can be calculated using Equation 4, proposed by Dix (1955), where V n and V n-1 are the average velocities from the soil surface to the top of reflectors n and n-1 and, respectively, and t n and t n-1 are the corresponding two-way travel times.
Several studies related the soil water content with the propagation velocity of an electromagnetic pulse.The first contributions to this subject used the Time Domain Reflectometry (TDR) technique (Conciani et al., 1996;Topp & Davis, 1981;Topp et al., 1980).Similar concepts can be applied to GPR (Amparo et al., 2007;Botelho et al., 2003;Machado et al., 2006), although the position of the reflectors has yet to be discovered, contrary to TDR.Equation 5 was proposed (Botelho et al., 2003) based on the Wyllie equation for the elastic wave velocity of propagation (Wyllie et al., 1958)   (5) Grain size distribution, mineralogy, porosity, and water content/salinity are the main parameters controlling the soil's electrical resistivity (ρ).Furthermore, the soil cation exchange capacity (CEC) can indicate the mobility of the ions around soil particles.These ions facilitate the electrical current flow through the soil.Soil clay content and the predominance of 2:1 minerals such as bentonite and montmorillonite increase CEC (Nguyen, 2014).Thus, coarser soils present higher resistivity than silty and clayey soils.Electrical resistivity tomography (ERT) provides continuous 2D images of the subsoil, making it possible to analyze the ρ variations laterally and with depth (Souza & Gandolfo, 2012).The analysis of the resistivity sections allows for identifying the resistivity anomalies due to different lithologies and water contents or contaminant plumes in the subsurface (Sass et al., 2008).
Although demanding the injection of higher current intensities due to its low signal to noise ratio, the dipoledipole arrangement allows rapid data acquisition and enables studying the lateral resistivity variations at different depths.
In such an arrangement, the current (AB) and the potential (MN) dipoles are placed/aligned on the ground surface.The distance between the electrodes is kept constant and equal to a.The data acquisition starts with a minimum distance x = 1  between the pairs AB, and MN and the following measurements are performed by displacing the pairs of electrodes at multiples of  (Souza & Gandolfo, 2012).Although providing continuous 2D and even 3D images of the subsoil, ERT resolution is a function of the distance a, and its use for high-resolution images is time-consuming.According to Zorzi & Rigoti (2011), the depth investigation can vary from AN/4 to AN/10.Furthermore, soil resistivity is a function of its water content, varying along the year according to the dry/rainy seasons.Using ERT associated with GPR offers a possibility to overcome these drawbacks.
In recent years, there has been an increase in the simultaneous use of GPR and ERT in site investigations, although such use has been limited to geotechnical applications.The purpose of site investigations varies depending on its objective, ranging from soil characterization and layer interface delimitation (Evangelista et al., 2017); to geological studies in unsaturated karst zones (Carrière et al., 2013), assessment of the thickness of talus layers in the European Alps (Sass et al., 2008), and stratigraphical characterization of Quaternary sediments layers (Pellicer & Gibson, 2011).
In such works, GPR data is used to better define the interface between layers or as an additional tool in delineating high

Study site
The study site is located at the countryside (thorp of Água Branca, 12°35'39.8"S;38°26'06.2"W) of the city of São Sebastião do Passé, Bahia, Brazil, about 68 km from the capital of Bahia.The local geology is dominated by the Todos os Santos sedimentary Bay, formed from the evolution of the crustal stretching, which caused the fragmentation of the Gondwana supercontinent in the Mesozoic era (Lima, 1999).
Sediments of the São Sebastião and Barreira Formations and Quaternary deposits predominate in the study area (Figure 1).The São Sebastião Formation is a fluvial deposit lithologically composed of fine to coarse sandstones interspersed with silty clay layers.The typical composition of the clayey layers presents kaolinite and illite with a considerable amount of iron oxides (Souza et al., 2004).
Due to its high effective porosity and hydraulic conductivity, this formation is one of the most important underground water reserves of Todos os Santos Bay (Alves, 2015;Lima, 1999).The Barreira Formation is formed by fine sand and kaolinitic silty clay fractions with crossbedding and plane-parallel lamination.Its thickness ranges from 30 m to 40 m (Ghignone, 1979).The alluvial Quaternary sediments occur in shallower depths in valleys, floodplains, and the coast (Barbosa & Dominguez, 1996).Lithologically, they are poorly graded sandy sediments usually rich in organic matter.According to Lima (1999), the Reconcavo aquifer system, composed of the São Sebastião, Marizal, and Barreiras Formations, has a water reserve estimation of approximately five hundred billion cubic meters and is used to attend villages, cities, and industrial facilities.

In-situ site investigation campaign
The investigation campaign (Figure 2) used five SPTs boreholes excavated down to 15 m depth (ABNT, 2020), three ERT, and two GPR sections (Farias, 2021).Investigation line 3 was positioned close to SPT boreholes.SPT results (Figure 3) were used to interpret the ERT data and define the electrical resistivity ranges in the ERT sections.Unfortunately, the SPT performed tests could not detect groundwater table.The cisterns provided the only evidence concerning the groundwater table position in some residences' backyards during the rainy season.
ERT surveys used a Syscal Pro resistivity meter (Iris Instruments ® ) with ten channels and an internal transmitter with 250 W and 2000 Vpp, allowing the injection of a maximum current of 2.5 A into the soil.The integrated transmitter/ receptor unities enable setting automatic reading scales and simultaneous measurements of apparent resistivity and chargeability.An inter-electrode spacing of  = 10 m was adopted (see Figure 4a).This value was adopted, aiming for a balance between ERT resolution and survey feasibility.The authors suggest a value of  = 5 m for better integration with GPR and SPT results for shallower investigations.A saline solution was used to improve the electrode/soil electrical conductivity.ERT surveys followed ABNT (2011).
The three ERT sections (ERT1, ERT2, and ERT3) had lengths of 270 m, 220 m, and 270 m, respectively.The experimental procedure initially positioned the two dipoles AB and MN in the first four electrodes and, while keeping the current dipole (AB) fixed, moving the potential dipole (MN) until a distance between the dipoles of 8.The procedure was then resumed by positioning the dipoles at electrodes 2 to 5 (offset from the last initial position of 1 a), and the procedure was repeated until the last possible position.
The ERT apparent resistivity was inverted using ZondRes2d © software.The inversion process adopted a maximum depth of 45 m and ten iterations.Based on the obtained results from ERT and SPT, the following correspondence between local lithology and electrical resistivity was assumed for the study area: Clayey silt (weathered shale): ρ < 80 Ωm; Silty sand (weathered sandstone): 80 Ωm ≤ ρ < 180 Ωm and Lateritic silty sand ρ ≥ 180 Ωm.Although shale layers, especially if saturated, usually present resistivity values less than 20 Ωm, in this case, the material is described as clayey silt with a considerable amount of fine sand.Similarly, the silty sand (weathered sandstone) layer has a significant amount of clay, which is likely responsible for its relatively low electrical resistivity values.The 2D ERT sections were georeferenced using a R © (2020) programming language script developed for this study, the local topography, and the location of the survey points.Similar procedures were adopted for GPR   data, in this case, using a different script.The GPR surveys used Mala Geoscience © equipment consisting of central unity (CUII), odometer, 25 MHz antennas, and laptop (see Figure 4b).A constant offset of 4 m between the antennas and a step distance of 0.5 m were adopted.The time window was 800 ns, corresponding to an investigation depth of about 40 m, if a V = 10 cm/ns is initially assumed for the soil.These values follow those usually suggested in the technical literature (Davis & Annan, 1989).The GPR data were processed using a R © (2020) script and the RGPR Libraries (Huber & Hans, 2018) with the following filter sequence: time zero correction -> dewow -> bandpass filter -> power time.The obtained radargrams were analyzed, and the existent reflection hyperbolas were fitted using Equation 1.This procedure enabled the creation of a velocity field for each radargram, which was used to convert travel time to depth (or elevation).Radargrams were then geo-referenced using the developed scripts.
The R © (2020) scripts developed for this study allows GPR and ERT data to be superimposed for better comparison and the 3D interpolation of the field data.The interpolated data can be used to construct 3D interactive images to visualize the investigation campaign results better.Fitted hyperbolas were also used to calculate the interval velocities according to Equation 4 and then estimate soil water content (Equation 5).In this case, values of G = 2.70 and e = 0.81 were assumed based on the data from undisturbed samples.

Results and analysis
Figure 5 presents the resistivity contours for section ERT3 jointly with the lithology data from SPT.There is a fair agreement between interpolated resistivity and soil lithology.However, the distinction between the clayey silt (weathered shale) and silty sand (weathered sandstone) layers was only sometimes satisfactory.These discrepancies are influenced by groundwater table position (SPT and ERT were performed in different periods) and the similar texture of the layers, besides ERT resolution, around 5 m.Individual bore log details can be found in Farias (2021).
Figure 6a presents data from the GPR2 radargram after filtering.Figure 6b shows some adjusted hyperbola and possible layer interfaces.The radargrams' hyperboles are valuable information often neglected in GPR surveys.They allow underground water content estimations using the TDR principles as previously discussed.Furthermore, the indicated layer interfaces help separate the different soil layers, adding resolution to the ERT surveys.Depth (z) and mean pulse velocity (V) are also indicated in Figure 6b (Equation 1). Figure 6c highlights high attenuation zones and crossbedding, which are coherent with the layering pattern observed in some SPT samples and with the findings of Lima (1999) concerning the Barreira Formation.High attenuation zones (loss of GPR return signal) in radargrams indicate the presence of high plasticity soils, which absorb most of the electrical-magnetic pulse energy due to their high electrical conductivity.
The high attenuation zones shown in Figure 6c were used with the estimated soil water contents to indicate the occurrence of the saturated shale layer.Equation 5 provided w estimations from V int , which were calculated using the values of V presented in Figure 6b (Equation 4).Higher values of V are observed on the left side of the radargram (Figure 6b), indicating less saturated soil for shallower depths in this region.Figure 7 presents soil water content contours for GPR2 and indicates the probable position of the groundwater table.The soil was considered saturated for pulse velocities V int ≤ 6.0 cm/ns, corresponding to w > 25%.This value follows the average value of void ratio of the collected specimens in the field (Farias, 2021).A sharp transition in the  values is noted on the left side of Figure 7, probably related to the occurrence of the sandstone (higher hydraulic conductivity) layer on the top of the shale layer.Therefore, rainwater is expected to infiltrate through sandstone and accumulate in the layers' interface.The groundwater table position obtained from GPR's surveys was coherent with the field observations performed in some cisterns located in the area during the rainy season.
Figure 8a presents the overlapping of ERT2 and GPR2 data.There is a good agreement between the high attenuation zones in GPR2 and the occurrence of the weathered shale layer indicated in ERT2.
The hyperbolas identified in the radargram appear to be related to the high-resistivity zones detected in the lower part of Figure 8a.Due to their position at great depths and below the water table, these data indicate that it is possible that it is a different lithology than the shallower high-resistivity layer (silty lateritic sand).Figure 8b presents the modified ERT2 section after the coupled analysis of the SPTs, electrical resistivity, and GPR campaigns.Figure 9a presents the overlapping of ERT1 and GPR1 data.In this case, the high attenuation zones extend beyond the areas initially marked    as the weathered shale layer, motivating the changes in the ERT interpretation, as indicated in Figure 9b.
Figure 10 shows some sections of the 3D model obtained by interpolating the electrical resistivity values.Downloading the following links provide interactive versions of the Figure 10: Sections 3D (Machado, 2023a) and Surfaces 3D (Machado, 2023b).

Conclusion
Results from different direct and indirect investigation methods are presented and discussed to highlight how they can be used in a coupled manner for better characterization of subsurface conditions.The GPR data provided valuable information concerning the high conductance zones and aided in delimitating the contacts between the different layers in the field.In addition, the pulse propagation velocities obtained from hyperbola fitting were fundamental in correcting the radargram depth, also making it possible to estimate the soil moisture distribution and groundwater table location.The estimated moisture content values provided valuable information to interpret the ERT data since water can widely change ER values for a given soil formation.The SPT provided the basis for sketching the subsurface model, correlating the data from direct and indirect methods, and providing the information for ERT ranges definition.The developed activities in this case study may contribute to better site characterization, aggregating data from different sources, and analyzing the results in a coupled and interactive manner.The investigation procedures reduce doubts about the geotechnical characteristics of the subsurface soil layers.Detection of the layer interface is improved, in addition to providing estimates of soil water content that helps define the groundwater table location.

Figure 2 .
Figure 2. In situ tests carried out at the study site.

Figure 7 .
Figure 7. Contours of estimated water content values for GPR2 section.
AbstractGeophysical methods are potent tools for geotechnical site characterization in a nondestructive way.They improve the extrapolation of punctual data from direct survey methods, allowing a fast and cost-effective evaluation of large areas.Ground Penetrating Radar (GPR) and DC electrical resistivity (ER) are the most requested methods for geotechnical and geoenvironmental applications.Their use, however, is usually uncoupled, with no sharing of information from one method to another to improve data interpretation.This case study illustrates the development of protocols and scripts in R © programming language for ER and GPR data analysis with Standard Penetration Tests (SPT) data to produce more accurate information on subsurface conditions concerning lithology, water content, and groundwater table (GWT) position.The SPT data were used to associate resistivity ranges with different soil lithologies and GPR pulse velocities for estimating the soil water content.Estimated water content values aided in interpreting ER data and locating the groundwater table.
).Most minerals and fluids present values of μ close to μ o .However, values of ε can vary widely in soils due to the soil water content.Furthermore, for the pulse
attenuation (or high conductivity) zones detected by ERT surveys.The use of GPR data to estimate moisture content and assist in interpreting ERT data is scarce.No publications with such characteristics were found in the literature by the authors.This case study uses SPT, GPR, and ERT for site characterization, including the water content estimation and definition of groundwater table position.SPT and GPR data are used to improve the ERT 2D sections, and the results of the modified ERT sections are used to construct a 3D stratigraphical model.