Interpretation of large‐scale, long‐term electrical geophysical monitoring guided by a process simulation

Surface electrical resistivity tomography (ERT) was used at a waste site to monitor vadose zone changes in electrical properties as a proxy for contaminant flux over a span of 17 years. The BC Cribs and Trenches (BCCT) site at the Hanford site contains 20 disposal trenches and six disposal cribs. Wastes include a large inventory of technetium‐99 and large masses of nitrate and uranium‐238. ERT data were collected along 41 profiles in 2005 to characterize regions of elevated bulk electrical conductivity (BEC) associated with past liquid waste discharges. Previous analyses performed on samples from four boreholes showed a high correlation between nitrate concentration and BEC. In 2022, ERT data were re‐collected along the same profiles and six additional profiles in an area not previously surveyed. Compared to background uncontaminated areas, BEC was higher in contaminated areas at the waste sites. Given the correlation between nitrate concentration and BEC previously found at this site, ERT images show the spatial distribution and relative ionic concentration of vadose zone contaminants at BCCT. Between 2005 and 2022, ERT difference images showed a decrease in BEC surrounding most waste sites, with exceptions where there were known anthropogenic surface changes. An evaluation of recharge‐driven nitrate migration using synthetic flow and transport simulations showed that downward migration causes a decrease in BEC from the decrease in ionic strength at the trailing end of the plume where contaminants migrated downward. From this, we interpret ERT difference images as showing the predominant regions of downward ion flux.

transport remain one of the foremost challenges and most critical aspects of site remediation.Understanding contaminant behavior in the vadose zone is essential because once the contaminants migrate out of the vadose zone into the groundwater, they can more rapidly reach and negatively impact areas with human, plant, and animal life.The contaminant flux from the vadose zone to the groundwater often persists over long time scales (e.g., decades and beyond) and has a Vadose Zone Journal direct relation to the spatiotemporal distribution of contaminants in groundwater (Oostrom et al., 2016).One way to understand contaminant flux is by conducting a geochemical and microbial analysis of cores, but this analysis is limited because the data are restricted to a single point source location and a snapshot in time.Recent sensor development aimed at in situ sampling of pore water and/or soil gases shows promise (Linneman et al., 2022), but these data are confined to borehole locations.There is a need to acquire spatially dense information to advance the understanding of contaminant flux in the vadose zone, for which a limited number of borehole samples are not representative of the whole subsurface environment.Geophysical methods can provide such datasets from either the surface or boreholes.These data can be modeled to provide an estimate of two-or three-dimensional (2D or 3D) geophysical parameters that are then correlated to subsurface physical parameters (Michot et al., 2003;Stan & Stan-Kłeczek, 2014;Sudha et al., 2009).
The Hanford site (Washington) is a decommissioned plutonium production complex with a deep vadose zone and significant quantities of subsurface hazardous waste.At the BC Cribs and Trenches (BCCT) site, a 30-ha waste disposal facility located on the Hanford site, approximately 30 million gallons of waste were received from uranium and ferrocyanide recovery processes from 1956 to 1958.Radiological and inorganic liquid waste was directly disposed into subsurface infiltration galleries, including 20 unlined cribs and six concrete-lined trenches (DOE/RL, 2008), because it was hypothesized that contaminants discharged to the subsurface could be selectively retained through ion exchange, with radioactivity decaying over time (Gee et al., 2007).Sev-

Core Ideas
• Surface ERT was used at a waste site to monitor vadose zone changes in electrical properties.• Time-lapse ERT was used as a qualitative measure of contaminant flux over a span of 17 years.• A flow and transport simulation was used to guide the interpretation of the ERT images.• Time-lapse ERT was able to detect predominant regions of downward ion flux.
eral borehole geophysical studies (Brodeur et al., 1993;Fecht et al., 1977;Horton & Randall, 2000) and geologic, geochemical, and hydraulic measurements (Serne et al., 2009) were conducted to assess that contaminants had not reached groundwater.They found that there was ongoing but limited vertical migration to approximately 70 m below ground surface but concluded that contaminants did not reach the water table, reported at a depth of 104 m (Benecke et al., 2006).
To supplement and further validate these studies, surface ERT data were collected to identify regions impacted by past discharges of liquid waste composed primarily of high sodium and nitrate concentrations (Benecke et al., 2006;Rucker & Fink, 2007).The ERT surveys consisted of a dense collection of 55 lines in a 2D orthogonal grid-like pattern on the surface of the site and took place in 2004, 2005, and 2006(Benecke et al., 2006;;Rucker & Fink, 2007).A qualitative 3D interpretation was made by interpolating 2D spatial interpretations of the ERT data, a strategy termed high-resolution resistivity (HRR).However, the authors acknowledged the limitations associated with this depth estimation method, where the 2D data analysis was translated to a 3D interpretation (Benecke et al., 2006).To address this shortcoming and solve for the true distribution of subsurface electrical conductivities, Rucker et al. (2009) performed a 3D inversion combining multiple 2D datasets.This is commonly referred to as a quasi-3D inversion (Cheng et al., 2019;Dahlin & Loke, 1997) and has been found to provide a better spatial interpretation relative to analyzing and interpolating between individual 2D datasets (Cheng et al., 2019;Papadopoulos et al., 2007;Robinson et al., 2015).Their results imaged a high bulk electrical conductivity (BEC) plume associated with nitrate at depth, but limitations on computer memory constrained the discretization of the modeling domain, resulting in a coarse image of the electrical structure of the subsurface.The authors tried to divide the site into smaller modeling domains to avoid exceeding computer memory, but the boundaries of the subdomains did not yield realistic results.Despite the limitations of the analysis, the results showed that contaminants had not yet reached the water table.In this study, we performed a repeat survey at BCCT to assess plume migration from 2005 to 2022 using time-lapse ERT difference imaging.In addition, six new profiles surrounding the BCCT cribs were acquired, which was an area that was not well covered by the original 41 profiles.Given the lack of baseline data, these additional profiles did not provide information on changes in BEC but were collected to identify where high BEC exists.Through supporting coupled contaminant transport and ERT simulations (Jaysaval et al., 2023), we demonstrate that ERT was able to locate the predominant regions of downward ion flux and could be a useful source of data to inform a flow and transport simulator.

SITE DESCRIPTION
BCCT is located south of the 200 East Area within the Hanford Central Plateau (Figure 1).The site received one of the largest liquid waste volumes released to the ground in the Hanford Central Plateau.The waste stream at BCCT began as uranium and fission products from the bismuth phosphate separation process, which were reprocessed at the 221-U Plant to recover uranium.After the uranium was removed, ferrocyanide was added to precipitate cesium and strontium.The supernate was released to BCCT based on a specific retention basis such that the liquid released was kept below a volume threshold so that contaminants would be held in the soil by surface tension and not released to the groundwater (McDonald et al., 2020).The radionuclide inventories at BCCT consist primarily of cesium (Cs-137), strontium (Sr-90), technetium-99 (Tc-99), and tritium (H-3) (Zaher & Agnew, 2018).In addition, there are high sodium and nitrate concentrations and potassium ferrocyanide that were used to complex/precipitate Sr-90 and Cs-137.The inventory of Tc-99 (∼410 Ci) represents the largest volume released at the Hanford site (DOE/RL, 2008).
While the trenches and cribs were both designed as liquid waste disposal sites, the cribs are smaller (∼12.2m 2 ) and designed to disperse liquid waste evenly.Trenches were designed to be long and narrow (∼150 m in length), and liquid waste was received through a network of pipes placed at infrequent intervals along the length of the trench, resulting in an uneven contaminant distribution (DOE/RL, 2008).
There are seven hydrological stratigraphic units in the vadose zone beneath BCCT, including (from the surface): backfill, Hanford formation (Hf) unit 1 (Hf1), Hf2, Hf3, the siltier end member of the upper Cold Creek unit (CCUsilt), the Ringold Formation member of Taylor Flat (Rtf), and the Ringold Formation member of Wooded Island-unit E (Rwie) (Figure 2).Hf (∼80-m thick) is the dominant sequence in the vadose zone beneath BCCT and consists of a thick sand-dominated sediment with a high (∼50%) basalt content (Martin, 2010).The differentiated Hf1, Hf2, and Hf3 units contain multiple beds of fine-to coarse-grained sand up to several meters thick.Within the 200 East Area, the Columbia River basalt has been identified in boreholes at depths >160 m.The water table elevation is approximately 121.5 m (Karanovic et al., 2021), which is about 106.5 m below ground surface.BEC is strongly dependent on soil temperature (Hayashi, 2004); however, these fluctuations become minimal (<1˚C) at depths >9 m (Hsieh et al., 1973).Thus, the spatial or temporal variations in BEC below a 9-m depth are not due to temperature fluctuations.
As part of the original 2004-2006 ERT study, four boreholes were drilled within BCCT to provide a link between surface-based electrical resistivity and direct measurements of geochemical, hydraulic, and soil resistivity (Serne et al., 2009).The boreholes were geologically logged in the field, and over 480 grab samples and a limited number of core samples were collected.The four boreholes contained large volumes of sand-dominated sediment relative to finer-grained silty material; however, the silty layers appeared to have a large impact on the distribution and lateral movement of moisture within the vadose zone (Serne et al., 2009).Neutronmoisture geophysical logs were also collected to compare to geological layers; these logs showed a good correlation between moisture content and fine-grained layers.
Data from one of the boreholes, C5923, are shown in Figure 3.In this borehole, the Hf is undifferentiated, but finer layers are denoted.There is an undifferentiated Hf/CCu (∼33-m thick) unit that consists of sand to gravelly sand with occasional layers of fine-grained silty sand with a 20%-40% basalt content.Pore fluid electrical conductivity (EC) was measured from grab samples using the supernatants from 1:1 sediment:water extracts.This shows a strong correlation between the EC and the high ionic content of nitrate, which is a co-contaminant of concern at this site.Moisture content is shown to be correlated with stratigraphy, which is more evident at shallower depths.While ERT cannot measure contaminant concentrations or flux directly, Figure 3 shows that BEC is correlated with nitrate concentration and therefore nitrate flux.

Field campaign details
Acquiring true 3D ERT data is time consuming and costly compared to 2D data acquisition (Rucker et al., 2009) and can also be impractical for larger areas with hundreds of electrodes (Cheng et al., 2019).Given these logistical challenges, the original 41 profiles of ERT data were collected as 2D transects.The profiles were organized as a grid to fill in areas that were not directly overlapping the cribs and trenches (Figure 4).Additionally, the ERT lines were extended past the crib and trench boundaries into adjacent properties to delineate the edges of the plume (Benecke et al., 2006;Rucker & Fink, 2007).Along each profile, most of the electrodes were spaced at 3 m, and several were spaced 6 m apart.
A repeat ERT data campaign was conducted in 2022 along the same 41 profiles originally collected in 2005 (Figure 4; orange electrodes).Compared to the original survey, every other electrode location was used, which resulted in 6-m electrode spacing along 39 profiles and 12-m electrode spacing along two profiles.Based on the study of T. Johnson et al. (2022), these spacings were deemed sufficient to provide the necessary resolution to monitor the deeper changes in BEC anticipated as a result of downward plume migration.In addition to the 41 repeated profiles, six additional profiles surrounding the cribs were collected (Figure 4; red electrodes).For these lines, a 6-m electrode spacing was used.Since there is no previous 2005 dataset against which to difference these additional profiles, the resulting images represent a static snapshot characterization of BEC.
Between 2005 and 2022, there were several localized surface changes at BCCT, which have the potential to influence BEC.Notably, a north-south trending road and gravel pad were installed after 2005 (Figure 4).The fill and surface cover of these new features change the topography and soil properties and can allow for additional moisture retention.In addition, there was a subsidence zone recently identified by the site contractor along 216-B-58 (Figure S1), which is along ERT Line 14 (Figure 4), which could also cause anomalous bulk conductivity compared to 2005 conditions.Between 2005 and 2022, we were unable to find any reported natural changes in surface vegetation that could impact BEC.

Electrical resistivity tomography
Electrical resistivity (the inverse of EC) quantifies how strongly a material opposes the flow of an electrical current.This is controlled by the porosity, moisture content, temperature, pore water fluid conductivity, soil texture, and mineralogy.ERT is an active source geophysical method that uses an array of electrodes to collect data.Surface ERT can also be referred to as electrical resistivity imaging (ERI).Electrode configurations with four electrodes (Ward, 1988) are typically used to inject a direct current into the subsurface using two electrodes; two other receiving electrodes are used to measure the voltage or potential difference (ΔV).The basic unit of ERT data is transfer resistance (Ω), which is the measured voltage drop (ΔV) across receiving electrodes divided by the injected current (I).ERT data can be represented as apparent BEC (Siemens/meter), which is computed based on analytical models depending on the configuration of electrode locations (Ward, 1988).Hundreds or thousands of ERT measurements can be collected in a single ERT survey.ERT data collected in 2005 used a pole-pole (PP) array (Ward, 1988).In this configuration, one current electrode (C1) and one potential electrode (P1) are placed at least 20 times the maximum separation distance between another current electrode (C2) and potential electrode (P2).C1 and P1 are referred to as remote electrodes (Figure 1).The PP array has a wider horizontal coverage and a deeper depth of investigation but a poorer resolution compared to more commonly used electrode configurations (Loke, 1999).At BCCT, the PP array was chosen for the 2005 data collection to provide rapid data acquisition (Rucker & Fink, 2007), as the geometry is highly efficient using multichannel instruments.
ERT data can also be collected using other dipole (DP) measurement configurations, including nested arrays (Wenner, Schumberger, and multiple-gradient) and DP-DP arrays (Ward, 1988).Small-offset DPs provide a high resolution in the vicinity of the electrodes and constrain near-surface structures in the ERT inversion.Large-offset DPs probe deeper into the subsurface.In addition to repeating the 2005 PP survey, a comprehensive combination of small-, intermediate-, and large-offset DPs was collected in 2022 to provide improved resolution for both shallow and deep structures to the extent possible.However, the DP-DP survey data cannot be analyzed using time-lapse difference imaging because there was no previous baseline dataset.
To distinguish between the two datasets, PP ERT data will be referred to herein as PP, and DP measurements as DP.Time-lapse ERT difference images will be shown for the PP data since they were collected in 2005; static ERT images will be shown for the DP data.For the PP and DP data acquisition, reciprocal measurements were collected to assess data quality, which is where the current and potential electrodes are swapped.Deviations between reciprocal measurements are considered a valuable way to characterize data noise (Tso et al., 2017).A full set of reciprocals was collected for 18 DP profiles and 22 PP profiles.Where time did not permit this collection, 10% of the full dataset was collected for 29 DP profiles and 25 PP profiles.The average reciprocal errors were typically less than 1% for each 2D dataset collected.Where field data quality was low, a repeat survey was performed.
While apparent BEC can give a better understanding of the overall data magnitudes and changes over time, these values cannot be used to spatially represent the distribution of BEC; numerical inversion is used for this purpose.The inversions were performed using E4D (T. C. Johnson, 2014; T. C. Johnson et al., 2010), which is a finite-element deterministic geophysical modeling and inversion code.To justify the zero-potential boundary conditions while maintaining computational efficiency, the mesh contains a more finely discretized foreground region surrounded by a more coarsely refined background region whose boundaries are a large distance (e.g., at least 100 times the length/width of the ERT survey area) from the nearest electrode.For this mesh, the lower boundary of the foreground region was the water table elevation of 121.5 m (Karanovic et al., 2021).

3.2.1
Pole-pole data processing PP data were collected along 41 profiles, and there were 2988 unique electrode positions (some profiles had overlapping electrode locations from perpendicular profiles).The accuracy of the ERT electrode positions for the 2005 ERT field campaign was reported to one-tenth of a meter.For the repeat survey in 2022, electrodes were surveyed with a differential global positioning system (GPS) having a reported accuracy of 1 mm.Where possible, the previously reported electrode locations were reestablished using the GPS; however, several field challenges, including the newly identified subsidence zone near 215-B-58, necessitated that several electrode locations be offset from their original location.In addition, the location of one remote electrode needed to be offset by ∼26 Vadose Zone Journal m because of site access issues.For the baseline and timelapse ERT processing of PP data, the GPS survey coordinates were used unless the electrodes needed to be repositioned as noted.In this case, the previously reported electrode locations were used in the baseline ERT processing.The mesh contained approximately 2.84 million tetrahedral elements, with surface topography defined on a 20 m by 20 m grid using Hanford Lidar (https://phoenix.pnnl.gov/phoenix/apps/gisexplorer/index.html).There were small, finite-surface elements where electrodes needed to be offset.The same 15,249 measurements were used in the baseline and time-lapse inversions to maintain the same survey sensitivity.The relative data errors were assumed to be 5%, and the absolute data error was assumed to be 0.01 Ω.An absolute data error was included so that very small measurements would not dominate the inversion.The time-lapse inversion (2022 data) used the baseline results from 2005 as a starting model and was constrained to be identical to the baseline image unless demanded otherwise by the data.Both the background and time-lapse data were fit to a  2 equal to 1, that is, the data were matched as well as they deserved to be based on the measurement quality assessed, as explained above.

Flow and transport process simulations
To evaluate the time-lapse changes in the ERT PP images due to recharge-driven contaminant migration, a flow and transport simulation was conducted using a PP survey along Line 37 (refer to Figure 4 for location) using PFLOTRAN (Hammond et al., 2012;Jaysaval et al., 2023).PFLOTRAN is a massively parallel subsurface flow and reactive transport code that can solve multiphase, multicomponent, and multiscale reactive flow and transport in porous materials.In this simulation, PFLOTRAN used flow and transport parameters such as the porosity and intrinsic permeability and discharged waste information to simulate hydrogeophysical variables, for example, liquid saturation and solute concentration.A petrophysical relationship using Archie's law (Archie, 1942) was then used to map these parameters and variables to BEC.All parameters were chosen to approximate the Hanford sediment properties reported in Oostrom (2022).PFLOTRAN simulated synthetic ERT responses along Line 37 as electrical resistance data.These data were then inverted using E4D to evaluate the ERT resolution and the associated changes that would be detected.The results from this PFLOTRAN simulation are used to aid in interpreting the BEC differences between 2005 and 2022.

3.2.2
Dipole data processing DP data were collected along 47 profiles (compared to 41 PP profiles), and there were 3491 unique electrode positions (some profiles had overlapping electrode locations from perpendicular profiles).Electrode positions were surveyed with a differential GPS system having a reported accuracy of 1 mm.Most electrode locations coincided with the previously reported 2005 locations, but there were a few deviations, as noted in the previous section.
A different mesh was generated for the DP data, which only included the electrodes used in the DP data collection.This mesh included the additional six ERT profiles and did not include any offset 2005 electrode locations (see Section 3.2.1).This resulted in a DP mesh with 2.34 million tetrahedral elements, which is approximately 500,000 less elements than the PP mesh.Surface topography was defined on a 20 m by 20 m grid using Hanford Lidar.The survey contained 262,407 measurements.The relative data errors were assumed to be 5%, and the absolute data error was dependent on the measurement configuration used to collect the data.Nested measurements were assumed to have an absolute error of 0.1 Ω; DP-DP measurements were assumed to have an absolute error of 0.01 Ω.The DP data were fit to a  2 equal to 1.

RESULTS
Plan view and cross-sectional ERT images are shown in this section for the PP and DP surveys.The images show the more finely discretized foreground mesh region only, which means that all images show the vadose zone to the depth of the water table.

Pole-pole
Prior to performing a time-lapse inversion, the raw data were reviewed to ensure changes were detected between 2022 and 2005.Figure 5 is a data histogram of the logarithmic difference in apparent BEC between 2022 and 2005.A positive logarithmic difference is representative of an increase in BEC from 2005 to 2022, while a negative difference represents a decrease in BEC.The kernel density estimate is also plotted to remove any bias from the number of bins used in the histogram plot.
The dataset showed large differences between 2005 and 2022, and the data are slightly skewed towards negative values.This plot was useful in (1) determining that a time-lapse inversion was advisable (e.g., if there were no changes in the data, then a time-lapse inversion would also not show changes) and (2) predetermining whether there would be both significant positive and negative BEC changes in the inversion.
ERT images for the PP data are shown in plan views in Figures 6 and 7, sliced at elevations of 225, 200, 180, 160, and  Per convention in displaying surface ERT images, these images have been clipped at a dip of 45˚on the outer edges, which are areas of relatively low resolution.The ERT images have a more refined BEC structure at higher elevations and become increasingly smeared with depth.This is due to the decrease in resolution with distance from the electrodes that is inherent in ERT imaging and the relatively small number of measurements informing the inversion at the edges of the imaged region.The depths in Figures 6 and 7 vary from the very shallow near surface to about 85 m below ground surface.
Noting that the most distant regions from the cribs and trenches represent native, unimpacted soil (i.e., the dark blue regions in the figures), several observations can be made from Figures 6 and 7. First, there are higher BECs beneath most trenches and cribs, except around 216-B-53A, 216-B-53B, 216-B-54, and 216-B-58 (refer to Figure S1).Second, there are several notable areas of increases in BEC from 2005 to 2022 due to surface changes, including a road and gravel pad that were installed and a newly identified subsidence area near 216-B-58 in the vicinity.The final region where there is a notable increase in BEC is near the southeasternmost crib 216-B-19 (refer to Figure S1).This increase is also near well 299-E13-20, which was decommissioned in 2020 with cement grout down to approximately 183 m (600 ft) below ground surface.For reference, well documents show a water table elevation of approximately 105 m (344 ft) below ground surface.If extensive fluids were released into the vadose zone during the decommissioning or cement grout was injected into the screen interval, a corresponding increase in BEC may have resulted.Well, documents do not contain this level of detail; therefore, we were unable to verify and/or identify the cause of this BEC increase.
Apart from anthropogenic shallow surface BEC increases and the BEC increases near well 299-E13-20, the most notable observation within the time-lapse results is the large spatial area over which there are smaller decreases in BEC.This decrease in BEC is consistently observed over most of the waste sites, yet there is still higher BEC (e.g., higher contamination below and adjacent to most of the waste sites).A decrease in BEC could potentially have been caused by lateral spreading, redistributing the higher BEC wastes over a larger area.However, we presume that from 2005 to 2022, lateral spreading would have been driven by capillary forces and therefore be minimal relative to the initial hydraulic loading from the disposal discharges.Therefore, we presume there was limited lateral spreading, and the difference images are not informative in this regard.We explore ERT resolution limitations and other possible explanations for decreases in BEC by incorporating the PFLOTRAN simulations below.
Select ERT cross-sections are shown along line profiles (Figures 8-12); since the measurements were collected along these profiles, this is where the image resolution is the highest in the quasi-3D inversion.Additional cross-sections can be found in the Supplemental Material.Like the plan views (Figures 6 and 7), the top and middle images show BEC from 2005 and 2022, respectively; the bottom images show the logarithmic difference from 2005 to 2022.Where the ERT lines overlapped, the cross-sectional view was extended, and these occurrences are noted with the line number and "extended." The images have been clipped on the sides at a dip of 45( where not crossing another profile), and the shaded area at the bottom is where the sensitivity of the ERT model to the data is presumed to be low based on the normalized sensitivities (T.C. Johnson & Wellman, 2015).
With depth, many of the cross-sections show (1) shallow increases (<10-m depth) in BEC and ( 2) deeper (>10-m depth) decreases in BEC.Some exceptions to these observations are the surface changes from the road and gravel pad installed after 2005 and increases in BEC near crib 216-B-19 and well 299-E13-20.

4.1.1
Flow and transport process simulations  Compared to the PFLOTRAN-simulated BEC distributions (a-e), the ERT images (b-f) illustrate the effects of the limited resolution in this case and offer insight on how to interpret field BCCT images.For example, the ERT data from PFLOTRAN are sensitive to plume migration, but this manifests in an ERT image only as a decrease in BEC that is smoothed to the surface on the upper, trailing edge of the plume (Figure 13e,f).The deeper increase in BEC on the leading edge of the plume, which is further from the surface electrodes, is not resolved.

Dipole
ERT plan view images for the DP data help to validate the PP images (Figures 6-12

DISCUSSION
The PP and DP characterization images reveal the extent of the high BEC.Capitalizing on the high correlation between total ionic content and electrical conductivity (Serne et al., ERT characterization images have been interpreted to the extent possible in the previous sections; therefore, we focus on the changes in BEC revealed in the time-lapse imaging.These images show a shallow (<10-m depth) increase in BEC with consistent deeper (>10 m) decreases in BEC.There are several exceptions to this statement, where there were increases in BEC with depth, and Section 3.1 outlined anthropogenetic surface and subsurface changes, which could account for these occurrences.We now focus on lines of evidence for processes affecting BEC.
Processes affecting BEC could result from changes to the pore water ion concentrations and moisture content over time, including (1) geochemical reactions such as precipitation of super-saturated contaminants within the pore water and (2) recharge-driven downward migration of contaminants over time.To assess the potential effects of geochemical reactions on BEC, geochemical equilibrium simulations were performed on pore water concentrations from three boreholes in shallow (1.5-2.5 m below ground surface) and deep (15-27 m below ground surface) sediments using data from Serne et al. (2009).Shallow sediments were dominated by high concentrations of phosphate and carbonate minerals, and these were expected to precipitate in the simulations.Shallow sediments also had elevated levels of sodium and nitrate; however, these were not expected to precipitate in the simulations.Therefore, at shallow depths (<10 m), the simulations suggest that there would be an overall decrease in pore fluid EC due to the large decrease in aqueous carbonate and phosphate.Deeper sediments (>10 m) were dominated by calcium, magnesium, sulfate, sodium, and nitrate, with little to no phosphate and/or carbonate minerals.Geochemical equilibrium simulations in the three boreholes showed no change in nitrate or sulfate.Therefore, precipitation of carbonate minerals could cause a decrease in the pore water EC at depth, but the change would be small because the very high nitrate and sulfate concentrations dominate the pore water EC.Overall, the precipitation changes predicted by the geochemical equilibrium simulations would likely result in small spatiotemporal decreases in BEC; however, the difference images are unlikely to show this subtle change.
Using the PFLOTRAN simulation (Figure 13) as an interpretation aid for recharge-driven migration for the field BCCT images, we may presume the deeper (>10-m depth) decreases in BEC indicate where contamination has migrated downward from 2005 to 2022.Shallow increases in BEC at many locations are likely caused by differences (increases) in the shallow subsurface saturation and/or soil temperature in 2022 compared to 2005.In some cases, these increases appear to extend below depths of 10 m, likely because of smoothing resulting from the limited resolution.In context of the PFLOTRAN simulation and corresponding inversion results in Figure 13, the 200 m and lower elevation difference images in Figures 6 and 7 show that a downward ion flux has occurred predominantly beneath the trenches (216-B-23 through 216-B-34, 215-B-20 through 216-B-22, and 216-B-52).This is consistent with the characterization images, which suggest the largest ion concentrations generally exist beneath the same trenches.The difference images at elevations of 160 and 140 m in Figure 7 suggest that deeper downward migration occurred beneath trenches (216-B-23 through 216-B-28).Although quantitative transport metrics (e.g., the downward migration velocity, the location of the center of mass, the anticipated future flux at the water table, etc.) cannot easily or confidently be derived from the ERT images alone, the resolution of the plume migration suggests that the ERT could be used to inform PFLOTRAN-based transport simulations to better constrain transport and fate predictions for BCCT.

CONCLUSION
ERT data were collected at BCCT in 2005 along 41 profiles.A repeat ERT survey was performed along the same 41 profiles to image changes in BEC over the 17-year period from 2005 to 2022.In 2022, new ERT surveys were also collected along six profiles surrounding the cribs to provide characterization images.The repeat surveys consisted of PP and DP ERT data, and these were used in quasi-3D inversions to show BEC distributions and time-lapse changes in BEC (for the original 41 profiles only).ERT results were presented as BEC images and time-lapse changes in BEC (for the original 41 profiles collected in 2005 only).BEC images allow the relative contrasts to be viewed to better understand where high BEC exists.At BCCT, high BEC was previously shown to be highly correlated with nitrate concentrations and high moisture content found in finer-grained layers (Figure 3).Within the limits of ERT imaging resolution, BEC images (for 2005 PP, 2022 PP, and 2022 DP) show the spatial distribution and relative moisture and ionic concentrations of vadose contaminants at BCCT.Beneath and surrounding most cribs and trenches, there exists a higher BEC below a 10-m depth.Exceptions to this general statement are beneath trenches 216-B-53A, 216-B-53B, 216-B-54, and 216-B-58; areas where ground surface changes occurred from 2005 to 2022 (e.g., road and gravel pad installations); and southeast of crib 216-B-19 near well 299-E13-20.The lateral and vertical footprints of elevated BEC are similar when comparing the 2022 PP to 2022 DP images and therefore validate the results from these independently processed datasets.
In contrast, time-lapse images help to better understand where the highest changes in BEC occurred and, therefore, how ions are migrating in the vadose zone.Presuming that there are no changes in the physical properties of the lithology, porosity, grain size sorting, and grain packing, changes in BEC can be correlated with changes in saturation and/or pore fluid conductivity subject to ERT resolution limitations.
Aside from the areas with changes to surface features noted above (which are not focused on plume migration), the timelapse ERT images beneath BCCT show shallow increases (<10-m depth) in BEC and deeper (>10-m depth) decreases in BEC.Supporting geochemical equilibrium simulations show to a first order that supersaturated contaminant precipitation would have a limited impact on BEC and ERT images in 2005 and 2022 (i.e., ∼70 years after disposal).Supporting joint PFLOTRAN and E4D recharge-driven contaminant migration simulations suggest that the ERT difference images are manifesting the downward displacement of the plume as a relatively diffuse decrease in BEC, resulting from the decrease in ionic concentration at the nearer-surface trailing end of the plume where contaminants migrated downward.Therefore, the ERT difference images are interpreted as showing the predominant regions of downward ion flux.Corresponding increases in BEC (caused by contaminant ions migrating into uncontaminated sediments) at the deeper, leading edge of the plume were not resolvable in the ERT images.The ERT images also revealed an extensive region where BEC increases at the southern end of the BCCT area near decommissioned well 299-E12-30, the cause of which has not been verified.

F
Location of BC Cribs and Trenches (BCCT) within the Hanford site central plateau.ERT, electrical resistivity tomography.With the advent of the distributed-memory parallel ERT inversion code E4D (T. C. Johnson, 2014; T. C. Johnson et al., 2010), T. Johnson et al. (2022) performed a quasi-3D inversion of the 2D lines collected at BCCT, combining ERT data collected in 2005 from 41 of the original profiles.E4D provided the capability to invert the BCCT dataset within a single model domain.Their proof-of-concept imaged a lateral footprint of the high BEC emanating beneath the cribs and most of the trenches.High BEC features were resolved to depths between 60 and 100 m.

F
I G U R E 3 Lithologic and grab sample data extracted for borehole C5923, adapted fromSerne et al. (2009).EC, electrical conductivity.

F
I G U R E 4 Electrical resistivity tomography (ERT) profiles collected in 2022.This included 41 lines, which were previously collected in 2005, and six newly collected lines surrounding the BC Cribs and Trenches (BCCT).The repeat survey electrodes are shown as orange circles, and the newly located electrodes are shown as red circles.The coordinate system shown is ESPG:2856, NAD83(HARN) Washington South.This coordinate system is used within all subsequent images.
140 m.Ground surface elevations range from approximately 225 to 231 m.The top row shows BEC images from F I G U R E 5 Logarithmic apparent bulk electrical conductivity differences between the 2005 and 2022 electrical resistivity tomography (ERT) datasets.KDE, kernel density estimate.the 2005 data, the middle row shows BEC images from the 2022 data, and the bottom row shows the logarithmic difference between the 2005 and 2022 images.As shown in the data plot in Figure 5, a positive logarithmic difference is representative of an increase in BEC from 2005 to 2022, while a negative difference represents a decrease in BEC.Viewing changes as a logarithmic difference allows for smaller changes to be better visible in the time-lapse difference images.

Figure 13
Figure 13 shows the PFLOTRAN-simulated BEC distribution and the inverted ERT images from 2005 (a and b) and 2022 (c and d) with the logarithmic changes (e and f).

F I G U R E 6
Plan views of pole-pole electrical resistivity tomography (PP ERT) images sliced at elevations of 225, 200, and 180 m.Images have been clipped at a dip of 45˚along the edges of the ERT profiles.Ground surface elevations range from approximately 225 to 231 m.BEC, bulk electrical conductivity.
) and are presented in FiguresS18-S19.The ERT cross-sections of the quasi-3D DP inversion provide new insights into the BEC along Lines 42-47 (Figure14).These are shown beneath the profiles where the resolution is highest in the quasi-3D inversion.The color scale is different than that in the plan views (FiguresS18-S19) to accentuate the BEC contrasts along these lines.An extended F I G U R E 7 Plan views of pole-pole electrical resistivity tomography (PP ERT) images sliced at elevations of 160 and 140 m.Images have been clipped at a dip of 45˚along the edges of the ERT profiles.Ground surface elevations range from approximately 225 to 231 m.BEC, bulk electrical conductivity.ERT cross-section of Line 8 is also shown to view the consistency between the BEC structure in the ERT images.The images have been clipped on the sides at a dip of 45˚(where not crossing another profile), and the shaded area at the bottom is where the sensitivity of the ERT model to the data is presumed to be low based on the normalized sensitivities (T.C.Johnson & Wellman, 2015).The highest BECs are observed along Lines 44 and 45, which are closest to cribs 216-B-16 and 216-B-19.We presume that low BECs (blue colors) along these profiles represent natural conditions without any contamination.

F
I G U R E 8 Pole-pole (PP) cross-sectional views showing an extended cross-section along Line 1. BEC, bulk electrical conductivity.F I G U R E 9 Pole-pole (PP) cross-sectional views showing Line 5. Decommissioned borehole 299-E13-20 is shown, which coincides with the location of a high increase in bulk electrical conductivity (BEC) from 2005 to 2002.2009), this can be translated as a first-order footprint of contamination.However, ERT images are limited in resolution, which is governed by many factors, including electrode spacing, distances from electrodes, field noise, measurement sequence, and BEC structure.Consequently, small-scale features may not be resolved, and large-scale features can appear as blurred or smeared BEC representations (T.J.Johnson et al., 2019).These limited-resolution effects are important to consider when interpreting ERT characterization images.Time-lapse ERT imaging can reduce some of the ambiguity inherent in characterization images by focusing on changes in BEC; however, the effects of limited resolution will persist in difference images for the same factors mentioned for ERT characterization.Pole-pole (PP) cross-sectional views showing an extended cross-section along Line 12. BEC, bulk electrical conductivity; Elev, elevation.F I G U R E 1 1 Pole-pole (PP) cross-sectional views showing an extended cross-section along Line 37. BEC, bulk electrical conductivity; Elev, elevation.

F
I G U R E 1 2 Pole-pole (PP) cross-sectional view along Line 26.BEC, bulk electrical conductivity; Elev, elevation.F I G U R E 1 3 PFLOTRAN input models (a-c) used to generate electrical resistivity tomography (ERT) data, which were used as input to E4D to generate the ERT images (d-f).The top images are in units of bulk electrical conductivity (BEC), while the bottom images are logarithmic difference images.

F
I G U R E 1 4 (a) Plan view image showing the locations of profiles 42-27 and the extension of Line 8. (b) Cross-sections oriented from southwest (SW) to northeast (NE).The dashed black line in (a) indicates the extent of the cross-sectional image for Line 8 in (b).BEC, bulk electrical conductivity.