Earthen dams subjected to variable hydrologic conditions are vulnerable to seepage and internal erosion, which can compromise structural integrity. Electrical Resistivity Tomography (ERT) is widely used to image moisture-driven resistivity anomalies associated with seepage, but field applications rarely allow quantitative evaluation of localization accuracy because seepage geometry is generally unknown. Here, we present a controlled laboratory benchmark that compares the seepage-path localization performance of two widely used ERT arrays—Schlumberger and Dipole–Dipole—using a homogeneous, laboratory-scale 3D earthen dam model with a prescribed seepage pathway of known geometry. ERT data were acquired along six 2D profiles spanning the crest and downstream slope and inverted consistently. Localization accuracy was quantified by comparing detected and ground-truth anomaly positions at three reference points: the anomaly-top depth (T) and the left and right lateral boundaries (L and R). Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) were computed across profiles to assess both typical error and sensitivity to larger deviations. The Schlumberger array consistently produced lower and more uniform localization errors than the Dipole–Dipole array. At point T, Schlumberger achieved MAE/RMSE of 0.67/0.71 cm, compared with 0.75/1.46 cm for Dipole–Dipole. For lateral boundaries, Schlumberger yielded MAE/RMSE of 1.50/1.68 cm at L and 0.50/0.65 cm at R, whereas Dipole–Dipole yielded 2.42/3.46 cm (L) and 1.67/2.06 cm (R). These results indicate that, under controlled conditions with known ground truth, the Schlumberger configuration provides more accurate and more consistent delineation of seepage-path depth and lateral extent than the Dipole–Dipole array. The study demonstrates the value of laboratory-scale physical modeling as a repeatable framework for array evaluation and offers quantitative guidance for designing seepage-focused ERT monitoring strategies for earthen dams.
Abrbekouh et al. (Wed,) studied this question.