Time-lapse Electrical Resistivity Tomography Data Research Articles

Mapping and monitoring dense non-aqueous phase liquid source zone by fused surface and cross-borehole electrical resistivity tomography

Effective characterization of dense non-aqueous phase liquid (DNAPL) source zones is crucial for remediating polluted sites. DNAPL often reside as residuals or pools within high-permeability lenses and above impermeable layers due to soil heterogeneity, gravity, and capillary barriers. Given the high cost of drilling, electrical resistivity tomography (ERT) techniques—including surface ERT and cross-borehole ERT, are commonly used for DNAPL source zone mapping and monitoring. However, the low spatial resolution of ERT increases uncertainty in source zone investigations. This study proposes a method for improving DNAPL mapping and monitoring by fusing surface and cross-borehole ERT data. Sandbox experiments were conducted to simulate a heterogeneous DNAPL source zone, employing both ERT methods for static mapping and dynamic monitoring. Reflective light imaging (RLM) was used to visualize DNAPL migration and provide saturation data, allowing for the quantification of ERT's effectiveness in characterizing DNAPL distribution. The results indicate that individual ERT methods face significant challenges in DNAPL source zone mapping due to background interference. Surface ERT alone tends to underestimate the extent of deeper DNAPL source zones. However, fusing surface and cross-borehole ERT results in a complementary enhancement of vertical spatial resolution, thereby improving the characterization of DNAPL source zones. The fusion of static and time-lapse ERT data substantially enhances DNAPL source zone mapping and monitoring capabilities. By calculating the ratio of the ERT-monitored area to the actual area using resistivity change contours (5 %, 10 %, 15 %), it was found that fusing surface and cross-borehole ERT data improved monitoring resolution by 50.48 % compared to surface ERT alone and by 22.95 % compared to cross-borehole ERT. Principal component analysis (PCA) was effective in fusing time-lapse data, while the weighted average method (WAM) outperformed PCA for static resistivity data fusion.

Deep learning to estimate permeability using geophysical data

Time-lapse electrical resistivity tomography (ERT) is a popular geophysical method to estimate three-dimensional (3D) permeability fields from electrical potential difference measurements. Traditional inversion and data assimilation methods are used to ingest this ERT data into hydrogeophysical models to estimate permeability. Due to ill-posedness and the curse of dimensionality, existing inversion strategies provide poor estimates and low resolution of the 3D permeability field. Recent advances in deep learning provide us with powerful algorithms to overcome this challenge. This paper presents a deep learning (DL) framework to estimate the 3D subsurface permeability from time-lapse ERT data. To test the feasibility of the proposed framework, we train DL-enabled inverse models on simulation data. Each measurement in both synthetic and field data is standardized by removing the mean and scaling the time-series to unit variance. This pre-processing step is necessary to bring simulation data closer to field observations. Subsurface process models based on hydrogeophysics are used to generate this synthetic data. Training performed on limited simulation data resulted in the DL model over-fitting. An advanced data augmentation based on mixup is implemented to generate additional training samples to overcome this issue. This mixup technique creates weakly labeled (low-fidelity) samples from strongly labeled (high-fidelity) data. The weakly labeled training data is then used to develop DL-enabled inverse models and reduce over-fitting. As both time-lapse ERT (1133048 features/realization) and 3D permeability (585453 features/realization) data samples are from a high-dimensional space, principal component analysis (PCA) is employed to reduce dimensionality. Encoded ERT and encoded permeability are generated using the trained PCA estimators. A deep neural network is then trained to map the encoded ERT to encoded permeability. This mixup training and unsupervised learning allowed us to build a fast and reasonably accurate DL-based inverse model under limited simulation data. Results show that proposed weak supervised learning can capture salient spatial features in the 3D permeability field. Quantitatively, the average mean squared error (in terms of the natural log) on the strongly labeled training, validation, and test datasets is less than 0.5. The R2-score (global metric) is greater than 0.75, and the percent error in each cell (local metric) is less than 10%. Finally, an added benefit in terms of computational cost is that the proposed DL-based inverse model is at least O(104) times faster than running a forward model once it is trained. Data generation, DL model training, and hyperparameter tuning to identify optimal neural network architectures utilized high-performance computing resources while the DL inference is performed on a standard laptop. Approximately, O(105) processor hours are used for generating data and DL tuning and training. We acknowledge that the data generation and DL model development are expensive. But once a DL model is trained, it can be re-used for inversion rapidly for the given system, with set physics and domain. Note that traditional inversion may require multiple forward model simulations (e.g., in the order of 10 to 1000), which are very expensive. This computational savings ≈O(105)−O(107) makes the proposed DL-based inverse model attractive for subsurface imaging and real-time ERT monitoring applications due to fast and yet reasonably accurate estimations of permeability field.

Strategies to overcome near surface disturbances while inverting time-lapse surface ERT data

An important basis for a reliable groundwater management is the detailed knowledge of the aquifer. We attempt here to determine groundwater flow velocity and direction of an hard rock aquifer in a groundwater protection area (Hastenrather Graben, Germany). In a common experimental set-up, we injected low salinity water in the aquifer and monitored the resistivity changes in the aquifer using time-lapse Electrical Resistivity Tomography (ERT). However, we witnessed strong influences from near surface resistivity fluctuations which were in the range of expected changes within the aquifer. We investigated these disturbances in more detail and determined two main sources of disturbance: a surface ditch, used for drainage with varying water fillings and, therefore, varying resistivities over time; secondly, changes of soil-atmosphere interactions, i.e., water accumulation in slight surface depressions after rain events. In addition to these main sources, the measured data is furthermore affected by random noise.In this paper, we compare three approaches to reduce the influence of these effects. After identifying disturbed areas, we exclude data near disturbed areas from the inversion as first approach; secondly, we refined the mesh at disturbed areas; and thirdly, we apply a region-based regularization to the time-lapse inversion.In a synthetic case study, all three approaches improve the inversion result. However, the region-based regularization recovers the input features best. The geometry of changes at depth are here less masked and the simulated injection body is better resolved. Nevertheless, the values are less accurately fitted than with a global regularization factor. Therefore, applying the approach on field data requires careful adjustment of the individual regularization factors to improve the time-lapse inversion results.We apply the combined first and third approach on our field data. As expected from the synthetic study, we see less masking and compensatory effects. The injection plume is visible in the resistivity distribution. The groundwater flow seems to be negligibly small, which is in accordance with an existing groundwater flow model of the aquifer.Our results confirm the fact that it is important to consider potential near-surface influences when analysing time-lapse ERT data. Based on a detailed synthetic case and an application to a aquifer injection study, we suggest to exclude data near heavily disturbed areas, in addition to applying a region-based regularization scheme.

Soil structure and soil moisture dynamics inferred from time-lapse electrical resistivity tomography

The semi-arid sagebrush steppe in the western United States faces pressures from the agriculture industry, recreation use, invasive grasses, and a changing climate. A key to facilitating the healthy management of this ecosystem is understanding the distribution and behavior of soil moisture in the vadose zone in both natural and agricultural settings. Within unsaturated environments, soil moisture is spatially and temporally heterogeneous, and changes in porosity and permeability within arid soils complicate characterization of soil hydrologic properties. Importantly, accumulations of ‘caliche’ or pedogenic calcium carbonate in arid soils can greatly limit permeability; however, observing the role that caliche plays in the hydrologic process is difficult because the installation of in situ instruments disturbs the soils and only provides information at a single point. To investigate vadose zone processes on a broad temporal and spatial scale, we installed a 7x8m2 electrical resistivity tomography (ERT) array at the Reynolds Creek Critical Zone Observatory in southwestern Idaho. Existing soil moisture observations show that infiltration is limited to depths less than 60 cm at this site, as compared to at least 90 cm at other sites in the watershed. To capture the seasonal wetting and drying of the soils, as well as the soils’ response to rainfall events, we monitored the site bi-weekly during the spring, summer, and fall of 2015 and 2016. A time-lapse ERT array was placed adjacent to coaxial impedance dielectric reflectometry (CIDR) probes so that the time-lapse ERT data could be referenced to precise measurements of volumetric water content. In addition to the measurements provided by the ERT array and CIDR probes, soil texture analysis and soil profile descriptions from a near-by soil pit show typical arid soil morphology, with accumulations of clays and calcium carbonate in the B horizon. The resulting ERT inversions show the following soil structure: (1) a high-resistivity top layer corresponding with minor amounts of pedogenic calcium carbonate; (2) a low-resistivity intermediate layer at depths corresponding with substantial accumulations (stage IV) of carbonate; and (3) a high-resistivity deep saprolite. The resistivity of the top layer varies seasonally with changes in precipitation, while the intermediate carbonate soil layer does not. This agrees well with the changes in soil moisture with depth measured by the CIDR probes and suggests that the top of the carbonate soil layer limits infiltration. However vertical structure and cracks within the carbonate soil layer create vertical preferential flow paths; resistivity within these flow paths responds to large precipitation events and seasonal changes in soil moisture. This implies that the preferential flow paths are a conduit for soil moisture flow that is not captured by the CIDR probes. From the combined interpretation of the ERT and CIDR we conclude that soil structure and the presence of calcic soil horizons inhibits soil moisture infiltration during both the summer dry months and the winter wet flux period; however, preferential flow paths provide an important vertical connection between the deep and shallow portions of the critical zone.

Integrated hydrogeophysical modelling and data assimilation for geoelectrical leak detection

Time-lapse electrical resistivity tomography (ERT) measurements provide indirectobservations of hydrological processes in the Earth's shallow subsurface at high spatial and temporal resolution. ERT has been used in the past decades to detect leaks and monitor the evolution of associated contaminant plumes. Specifically, inverted resistivity images allow visualization of the dynamic changes in the structure of the plume. However, existing methods do not allow the direct estimation of leak parameters (e.g. leak rate, location, etc.) and their uncertainties. We propose an ensemble-based data assimilation framework that evaluates proposed hydrological models against observed time-lapse ERT measurements without directly inverting for the resistivities. Each proposed hydrological model is run through the parallel coupled hydro-geophysical simulation code PFLOTRAN-E4D to obtain simulated ERT measurements. The ensemble of model proposals is then updated using an iterative ensemble smoother. We demonstrate the proposed framework on synthetic and field ERT data from controlled tracer injection experiments. Our results show that the approach allows joint identification of contaminant source location, initial release time, and solute loading from the cross-borehole time-lapse ERT data, alongside with an assessment of uncertainties in these estimates. We demonstrate a reduction in site-wide uncertainty by comparing the prior and posterior plume mass discharges at a selected image plane. This framework is particularly attractive to sites that have previously undergone extensive geological investigation (e.g., nuclear sites). It is well suited to complement ERT imaging and we discuss practical issues in its application to field problems.

Coupled hydrogeophysical inversion of DNAPL source zone architecture and permeability field in a 3D heterogeneous sandbox by assimilation time-lapse cross-borehole electrical resistivity data via ensemble Kalman filtering

Characterization of dense non-aqueous phase liquid (DNAPL) distribution is important to facilitate the decision of remediation strategies. However, it is still a great challenge to characterize DNAPL source zone architecture with high resolution due to subsurface heterogeneity and relatively sparse data from traditional hydrogeological investigations. To overcome difficulties from such sparse data, electrical resistivity tomography (ERT) is introduced to locate DNAPL using time-lapse cross-borehole measurements. Due to the significant impact of geological heterogeneity on DNAPL source zone architecture, a data assimilation framework based on the coupled multiphase fluids-ERT model is developed to jointly invert DNAPL saturation and the permeability field using time-lapse ERT data. To validate the efficiency and performance of this framework, synthetic and laboratory experiments are both performed to monitor DNAPL migration and distribution in 3D heterogeneous sandbox with cross-borehole ERT. Result shows that time-lapse ERT and direct inversion can map the evolution of the DNAPL plume but loses details regarding the plume morphology due to the over-smoothing caused by geophysical inversion using an isotropic and homogeneous roughness-based regularization procedure. By contrast, the coupled inversion is successful to characterize both the permeability field and the evolution of the DNAPL plume with a higher resolution. This is because the coupled inversion is able to directly translate raw geophysical data into hydrologic meaningful information and therefore avoid artifacts caused by direct geophysical inversion.

Improvement of 2D ERT measurements conducted along a small earth-filled dyke using 3D topographic data and 3D computation of geometric factors

In the framework of earth-filled dykes characterization and monitoring, Electrical Resistivity Tomography (ERT) turns out to be a commonly used method. 2D sections are generally acquired along the dyke crest thus putting forward the question of 3D artefacts in the inversion process. This paper proposes a methodology based on 3D direct numerical simulations of the ERT acquisition using a realistic topography of the study site. It allows computing ad hoc geometrical factors which can be used for the inversion of experimental ERT data. The method is first evaluated on a set of synthetic dyke configurations. Then, it is applied to experimental static and time-lapse ERT data set acquired before and after repair works carried out on a leaking zone of an earth-filled canal dyke in the centre of France. The computed geometric factors are lower than the analytic geometric factors in a range between −8% and − 18% for measurements conducted on the crest of the dyke. They exhibit a maximum under-estimation for intermediate electrode spacings in the Wenner and Schlumberger configurations. In the same way, for measurements conducted on the mid-slope of the dyke, the computed geometric factors are higher for short electrode spacings (+18%) and lower for lower for large electrode spacings (−8%). The 2D inversion of the synthetic data with these computed geometric factors provides a significant improvement of the agreement with the original resistivity. Two experimental profiles conducted on the same portion of the dyke but at different elevations also reveal a better agreement using this methodology. The comparison with apparent resistivity from EM31 profiling along the stretch of the dyke also supports this evidence. In the same way, some spurious effects which affected the time-lapse data were removed and improved the global readability of the time-lapse resistivity sections. The benefit on the structural interpretation of ERT images remains moderate but allows a better delineation of the repair work location. Therefore, and even if the 2D assumption cannot be considered valid in such a context, the proposed methodology could be applied easily to any dyke or strongly 3D-shaped structure using a realistic topographic model. It appears suitable for practical application.

Forward modeling to investigate inversion artifacts resulting from time-lapse electrical resistivity tomography during rainfall simulations

Time-lapse electrical resistivity tomography (ERT) is commonly used as a minimally invasive tool to study infiltration processes. In 2014, we conducted field studies coupling variable intensity rainfall simulation with high-resolution ERT to study the real-time partitioning of rainfall into surface and subsurface response. The significant contrast in resistivity in the subsurface from large changes in subsurface moisture resulted in artifacts during the inversion process of the time-lapse ERT data collected using a dipole-dipole electrode array. These artifacts, which are not representative of real subsurface moisture dynamics, have been shown to arise during time-lapse inversion of ERT data and may be subject to misinterpretation. Forward modeling of the infiltration process post field experiments using a two-layer system (saprolite overlain by a soil layer) was used to generate synthetic datasets. The synthetic data were used to investigate the influence of both changes in volumetric moisture content and electrode configuration on the development of the artifacts identified in the field datasets. For the dipole-dipole array, we found that a decrease in the resistivity of the bottom layer by 67% resulted in a 50% reduction in artifact development. Artifacts for the seven additional array configurations tested, ranged from a 19% increase in artifact development (using an extended dipole-dipole array) to as much as a 96% decrease in artifact development (using a wenner-alpha array), compared to that of the dipole-dipole array. Moreover, these arrays varied in their ability to accurately delineate the infiltration front. Model results showed that the modified pole-dipole array was able to accurately image the infiltration zone and presented fewer artifacts for our experiments. In this study, we identify an optimal array type for imaging rainfall-infiltration dynamics that reduces artifacts. The influence of moisture contrast between the infiltrating water and the bulk subsurface material was characterized and shown to be a major factor in contributing to artifact development. Through forward modeling, this study highlights the importance of considering array type and subsurface moisture conditions when using time-lapse resistivity to obtain reliable estimates of vadose zone flow processes during rainfall-infiltration events.

Covariance-constrained difference inversion of time-lapse electrical resistivity tomography data

Hydrogeophysics has become a major field of research in the past two decades, and time-lapse electrical resistivity tomography (ERT) is one of the most popular techniques to monitor passive and active processes in shallow subsurface reservoirs. Time-lapse inversion schemes have been developed to refine inversion results, but they mostly still rely on a spatial regularization procedure based on the standard smoothness constraint. We have applied a covariance-based regularization operator to the time-lapse ERT inverse problem. We first evaluated the method for surface and crosshole ERT with two synthetic cases and compared the results with the smoothness-constrained inversion (SCI). These tests showed that the covariance-constrained inversion (CCI) better images the target in terms of shape and amplitude. Although more important in low-sensitivity zones, we have observed improvements everywhere in the tomograms. Those synthetic examples also show that an error made in the range or in the type of the variogram model had a limited impact on the resulting image, which still remained better than SCI. We then applied the method to cross-borehole ERT field data from a heat-tracing experiment, in which the comparison with direct measurements showed a strong improvement of the breakthrough curves retrieved from ERT. This method could be extended to the time dimension, which would allow the use of CCI in 4D inversion schemes.

Estimation of unsaturated hydraulic parameters in sandstone using electrical resistivity tomography under a water injection test

Hydraulic conductivity is an important soil property when determining the potential for water movement in topsoil and in spite of its importance; soil hydraulic conductivity remains one of the most difficult of soil properties to assess. Laboratory methods have limitations due to the size of the samples and taking undisturbed soil samples is usually difficult in sandy soil and in-situ methods are required to estimate hydraulic conductivity.This study was conducted to estimate saturated hydraulic conductivity in unsaturated sandstone using the ground surface electrical resistivity tomography (ERT). The site is characterized by a deep Arenosol soil with high permeability and a low water retention capacity located at the Semora-Correia, the east of Lisbon. Eight ERT snapshots were collected during a water injection test to produce a sequence of 2D resistivity images. Time-lapse ERT data were inverted using independent data inversion, the difference inversion and simultaneous space–time inversion methods. Afterward, using an in-situ approach resistivity variation models were converted to water content images. By comparing first and second spatial moments of water movement images inferred from the ERT method with unsaturated flow simulation predicted from a numerical solution of Richards' equation, the range of saturated hydraulic conductivity is estimated to be in 0.5–0.7 (cm/min).The evaluation of ERT approach was made using a synthetic test. The results of synthetic test showed that the estimated parameters were significantly influenced by the ERT inversion method and an overprediction of spatial moments and consequently saturated hydraulic conductivity was observed in all inversion methods; however the resistivity models obtained by simultaneous space–time inversion method was more successful in water movement monitoring.

Monitoring the saltwater intrusion by time lapse electrical resistivity tomography: The Chioggia test site (Venice Lagoon, Italy)

Abstract A novel experiment of time lapse electrical resistivity tomography (TL-ERT) aimed at monitoring the dynamics of the saltwater intrusion in the coastland bounding the southern Venice Lagoon is presented. A dedicated apparatus was developed and operated for about 9 months from November 2005. The system acquired ten resistivity tomograms per day, five of which with high resolution by a 97.5 m long and 2.5 m electrode spacing ERT line, and five by a 300 m long and 5 m electrode spacing line down to 50–60 m depth. The stratigraphy of a 50 m deep borehole drilled in the nearby of the ERT-TL alignment outlines the presence of a shallow phreatic aquifer in the shallower 12 m thick unit, followed by a semi-confined aquifer between 18 and 38 m depth and a locally confined aquifer down to the bottom. The shallow aquifer is the most contaminated by the salt intrusion with a minimum value of the formation resistivity equal to 1.0 ohm m corresponding to a salinity of 25–30 gr/l. A seasonal resistivity fluctuation is observed, with the saltwater front that intrudes landward during the autumn–winter season and moves back seaward in spring–summer. The first semi-confined aquifer is characterized by resistivity value of about 5 ohm m, while the confined aquifer is less contaminated showing resistivity values greater than 7.5 ohm m. For both the two confined aquifers the resistivity value rises at the beginning of the summer probably due to the seasonal fresh water recharge supplied regionally from the mainland. The TL-ERT data have been correlated with a number of environmental variables. A relationship is found between the resistivity in the upper 3–4 m of the phreatic aquifer and the rainfalls, and between the water level in the adjacent main channel and the resistivity down to about 10 m depth. With respect to the tidal regime, a daily/weekly correlation with resistivity changes is not evidenced, while a significant negative correlation coefficient exists at monthly to seasonal time scale for depths below 15 m. This experience suggests that TL-ERT is a powerful technique to perform multi-scale contaminant monitoring at different time scales.

Application of time-lapse ERT imaging to watershed characterization

Time-lapse electrical resistivity tomography (ERT) has many practical applications to the study of subsurface properties and processes. When inverting time-lapse ERT data, it is useful to proceed beyond straightforward inversion of data differences and take advantage of the time-lapse nature of the data. We assess various approaches for inverting and interpreting time-lapse ERT data and determine that two approaches work well. The first approach is model subtraction after separate inversion of the data from two time periods, and the second approach is to use the inverted model from a base data set as the reference model or prior information for subsequent time periods. We prefer this second approach. Data inversion methodology should be consideredwhen designing data acquisition; i.e., to utilize the second approach, it is important to collect one or more data sets for which the bulk of the subsurface is in a background or relatively unperturbed state. A third and commonly used approach to time-lapse inversion, inverting the difference between two data sets, localizes the regions of the model in which change has occurred; however, varying noise levels between the two data sets can be problematic. To further assess the various time-lapse inversion approaches, we acquired field data from a catchment within the Dry Creek Experimental Watershed near Boise, Idaho, U.S.A. We combined the complimentary information from individual static ERT inversions, time-lapse ERT images, and available hydrologic data in a robust interpretation scheme to aid in quantifying seasonal variations in subsurface moisture content.