3D effects on 2D resistivity monitoring in earth‐fill dams

Abstract

Measuring resistivity is a potentially powerful method of monitoring leakage zones that have developed in a dam, and their expansion over time. Generally, for embankment dams, two-dimensional (2D) resistivity data have been measured along the dam crest for the detection of leakage zones. However, the three-dimensional (3D) effects created by specific dam geometry and fluctuations in reservoir water levels significantly distort the 2D resistivity data measured at the dam crest. This study evaluates the 3D effects through 3D resistivity modelling software, which was developed to calculate apparent resistivity data for geometries and material distributions for embankment dams. These modelling results demonstrated that the 3D effect from the dam geometry and variations in water level is significant. Especially, in the case of monitoring, changes in 3D effects from water level fluctuations cause a spurious near-surface layer when time-lapse inversion is applied with a cross-model constraint. To overcome this problem, we introduced a combined reference model constructed from the independent inversion of both time-lapse data and original reference data. The combined reference model was able to effectively suppress the spurious near-surface layer and to clearly image the damaged zone when the change in water level was small. However, a time-lapse inversion using the combined reference model also failed to identify the damaged zone when the change in water level was large. Finally, by using the resistivity monitoring system devised for dam surveillance to a test dam site, resistivity monitoring data were acquired. From the time-lapse inversion of two data sets showing a large change in water level between two measurements, it was confirmed that the variation of water levels produces the occurrence of a spurious near-surface layer due to a strong 3D effect. remain in time-lapse inversion of resistivity monitoring data for more detailed and quantitative interpretation.Internal erosion is an inevitable by-product of ageing for all embankment dams. This internal erosion generally develops into piping over a long time through a combination of backward erosion and concentrated leaks, a process that eventually leads to dam failure. Thus internal erosion and piping are major causes of embankment dam failure. Internal erosion initially results in increased porosity due to the loss of fine particles in the core. Resistivity is known to be very sensitive to changes in porosity in embankment dams. Especially in embankment dams with clay cores, the resistivity of a damaged core increases since saturated clay generally exhibits lower resistivity than that of fresh ground water. On the other hand, the saturated downstream shell shows a decrease in resistivity with the increase of water and clay content. The expansion of the leakage zone into surrounding areas results in an increase in leakage, leading to changes in resistivity.

A cost‐effective 3D electrical resistivity imaging approach applied to dike investigation

“…On one hand, their inner structure may exhibit strong longitudinal and transverse variability (e.g., presence of a clay core or repaired zones). On the other hand, embankment topography and high water‐level fluctuations adjacent to the dike have 3D effects on the resistivity measurements leading to image artefacts (Hennig, Weller, and Canh ; Sjödahl, Dahlin, and Zhou ; Oh ; Fargier et al ; Cho et al ). In order to account for the full 3D attributes of the problem, one would normally recommend the use of a complete 3D ERI procedure (e.g., Chambers et al ).…”

“…The addition of regularisation terms in inverse problem formulations is a powerful means of introducing prior information that has been proved to be essential for retrieving reliable solutions (Ellis and Oldenburg ). This has led to a variety of complementary techniques such as adding a smoothness constraint on the model to avoid unnecessary structure (e.g., Constable, Parker, and Constable ; deGroot‐Hedlin and Constable ), constraining the inverted model to a reference model (e.g., Oldenburg and Li ; Pidlisecky, Haber, and Knight ; Catt, West, and Clark ; Caterina et al ), using variable weighting factors depending on the reliability of prior information (Kim et al ), adding structural information such as known boundaries in the model (e.g., Kaipio et al ; Caterina et al ), decoupling the regularisation effects to preserve sharp boundaries where they are known to exist (Coscia et al ), or imposing bounds to the model resistivity values in selected regions by means of inequality constraints (Kim, Song, and Lee ; Cardarelli and Fischanger ). In the presented work, we implemented various modes of integrating a priori information, either by means of regularisation (as in all the previous references) or, alternatively, within the model and parameter mesh generation phase before the inversion process (Günther and Rücker ; Cardarelli and Fischanger ).…”

Applying Electrical Resistivity Tomography in Ornamental Stone Mining: Challenges and Solutions

“…The topography of the survey area and the exact electrode location is usually included in the data inversion, as they are known to have a significant impact on the measured data [43][44][45][46][47][48]. It has also been shown that topographic features outside the actual survey area can impact upon the measured resistivity data and can lead to misinterpretation, especially in the case of 2D data analysis [49,50]. In the study presented here, the survey area is located within a 30 × 40 m excavation that is 10-35 m deep (Figure 3).…”

Resistivity imaging of river embankments: 3D effects due to varying water levels in tidal rivers

“…The 2.5D inversion method (following references to 2D inversion imply the 2.5D assumption) assumes that the resistivity does not vary in the direction perpendicular to the vertical plane below the line. The perpendicular topographic variations of the embankment and chang-ing water levels to the side violate this assumption (Cho et al, 2014). As such, the data acquired from a 2D survey may be influenced by features adjacent to the survey, for example.…”