Quantifying seepage dynamics in a real-scale levee using time-lapse 3D ERT and novel spatiotemporal metrics

Aging levees face high failure risks due to undetected subsurface defects and undocumented construction, which complicates condition assessments. In this study, time-lapse 3D electrical resistivity tomography (ERT) was integrated with geotechnical and piezometric data to quantitatively assess the seepage dynamics in a real-scale levee under controlled flood conditions over 15 days. Four novel metrics were introduced: the detection rate (DR), saturation advance rate (SAR), total saturation period (TSP), and leakage persistence index (LPI). Results revealed two distinct seepage patterns across the levee sections. The levee crest showed rapid preferential wetting (DR = ∼1.34 m/day) and persistent saturation (LPI = 5.00–8.00). In contrast, the protected side exhibited delayed ingress (∼12 days), slower advance (DR ∼0.79 m/day), and prolonged saturation (TSP up to 15 days, LPI = 1.88). The ERT-derived water levels were highly precise, with a mean absolute percentage error (MAPE) of ≤1.88 % compared to piezometer data. This study presents a reproducible metric-based framework that transforms ERT from qualitative imaging into an actionable tool for early warning and proactive flood-risk management in levees.