Application of Vertical-Array Lateral Scanning in Seepage Detection of Urban Levees with Adjacent Underground Spaces

Abstract

With the increasing development of underground spaces adjacent to urban levees, contact seepage frequently occurs at the interface between the soil and underground structures. However, traditional geophysical detection methods are often rendered ineffective in such environments due to spatial restrictions and detection blind spots. To address these challenges, this paper proposes a vertical-array lateral scanning detection method. This approach utilizes electrical resistivity tomography (ERT) with flat-base electrodes and ground-penetrating radar (GPR) to acquire data directly from vertical wall surfaces. The feasibility of this method is validated through numerical simulations and field data. The results indicate that the proposed method effectively overcomes the high-resistance shielding effect of hardened walls and clearly reveals the electrical structure of the soil behind the wall. Specifically, the contact seepage zone manifests as a layered low-resistivity feature immediately adjacent to the wall, while the penetrating leakage channel presents as a continuous low-resistivity anomaly extending from the contact interface deep into the levee body. These findings confirm the applicability of this technology for the qualitative identification and effective detection of hazards in complex, space-restricted urban environments.

1. Introduction

Flooding poses a prominent issue in China, and flood control is directly related to public safety and socioeconomic development. Levee systems constitute a fundamental component of the flood mitigation infrastructure. Most levees in China were constructed between the 1950s and 1970s, and many are historical levees that have been repeatedly repaired and reinforced over time [1]. Owing to the limitations of construction technology and economic conditions at that time, these levees commonly exhibit inherent deficiencies, such as low design standards and uneven quality control. With increasing service duration, and under the long-term effects of hydraulic erosion, environmental weathering, and human activities, levees inevitably develop defects such as seepage, piping, cracks, and animal burrows, leaving some levees in a long-term state of “diseased operation” [2,3].

Cities are characterized by high levels of economic activity, dense populations, and concentrated fixed assets. Approximately 90% of large and medium-sized cities in China are located along rivers, lakes, or coastlines [4]. With accelerating urbanization, urban hydraulic engineering is facing increasingly complex risk challenges [5,6].

A large number of “levee-adjacent” and “levee-crossing” construction projects have emerged, including waterfront platforms, underground commercial spaces, tunnels, and landscape facilities. While these projects enhance urban functions, they inevitably disturb the original levee structure. In particular, the interfaces between newly built concrete structures and existing earth or rockfill levees often become weak links in the seepage control system [7]. Once contact erosion or seepage occurs, its concealed nature and location at critical flood-control sections make it prone to rapidly developing into piping failures, leading to serious economic losses. For example, during the historically high water levels of the Mississippi River in 2009, a levee breach occurred in the downtown area of Davenport, Iowa, causing floodwaters to inundate large numbers of shops and office buildings. Therefore, in such complex urban environments, the development of rapid, non-destructive, and high-resolution geophysical detection techniques for levee hazards is of great engineering significance and social value for identifying the spatial distribution of defects and supporting precise mitigation measures.

Geophysical exploration techniques, owing to their non-destructive nature, high efficiency, and high resolution, have become the mainstream approaches for detecting seepage in levees.

In terms of methodological applications, electrical resistivity tomography (ERT) and ground-penetrating radar (GPR) are currently the most widely used techniques. Binley et al. (2015) systematically reviewed the application of ERT in hydrogeology and pointed out its high sensitivity to groundwater flow and moisture variation, making it a preferred method for identifying seepage channels in embankments [8]. Cardarelli et al. (2014) combined ERT with seismic wave tomography to successfully reveal the anomalous distribution and structural features within an earth-filled dam. Regarding the precise localization of water seepage [9]. Demirci et al. (2012) utilized GPR combined with a back-projection algorithm to achieve high-precision imaging of water leaks from buried pipelines [10]. Cataldo et al. (2014) systematically compared the application effectiveness of GPR, ERT, and Time Domain Reflectometry (TDR) in underground leak detection [11]. In recent years, multiphysics-integrated detection has become a major trend. By combining the depth penetration advantage of ERT with the high-resolution capability of GPR, the ambiguity inherent in single-method interpretation can be effectively reduced [12,13,14,15,16].

However, although existing techniques have become relatively mature for conventional earth and rockfill levees, they exhibit significant limitations in dense urban environments with adjacent underground spaces. These limitations are mainly reflected in the following three aspects.

(1)

Geometric constraints of observation space: Traditional geophysical surveys typically arrange survey lines on open areas such as levee crests or slopes and adopt a top-down “half-space” detection mode. In urban levee sections, however, the crest and the top of adjacent underground spaces are often obstructed by buildings or green belts, making it impossible to deploy continuous survey lines. As a result, conventional methods face the dilemma of having “no place to measure.”

(2)

Coupling difficulties on hardened surfaces: Underground spaces adjacent to levees are mostly supported by vertical concrete retaining walls. On high-resistivity hardened surfaces such as concrete, the steel spike electrodes required by conventional resistivity methods cannot be driven in, resulting in excessively high contact resistance, insufficient injected current, and a severe reduction in data signal-to-noise ratio.

(3)

Detection blind zones at interfaces: The interface between underground structures and levee soils is a weak zone of levees and also a high-incidence area for contact seepage. This interface is usually vertical to the ground surface, and its lateral extent is often smaller than its vertical extent. When conventional surface-based methods are used, current lines or electromagnetic waves mainly propagate downward along stratification, making it difficult to capture the lateral heterogeneity of the interface. This geometric “misalignment” causes interface-related hazards to become blind zones in existing detection systems.

In summary, studies on lateral detection techniques for vertical hardened surfaces under spatially constrained conditions remain scarce. There is an urgent need to establish a specialized detection system suitable for urban levees with adjacent underground spaces.

This study proposes a vertical-array lateral scanning detection method for underground spaces adjacent to levees. This method arranges ERT and GPR measuring arrays along vertical retaining walls and implements lateral scanning data acquisition, which differs from conventional ground-layout geophysical techniques. It effectively solves the problems of limited deployment space and detection blind zones in dense urban environments, and realizes collaborative detection of shallow structural defects and deep seepage channels by integrating the advantages of GPR high-resolution imaging and ERT deep resistivity inversion.

2. Materials and Methods

2.1. Geophysical Methods and Principles

Common techniques for detecting levee hazards include electrical methods, electromagnetic methods, seismic methods, quasi-flow-field methods, and isotope tracing, with instruments such as ground-penetrating radar, transient electromagnetic systems, nodal seismographs, array resistivity systems, and piping–seepage detectors being widely used [17,18,19,20,21,22,23,24]. However, under lateral detection scenarios in underground spaces, transient electromagnetic methods suffer from shallow blind zones, and the transmitting and receiving coils are difficult to move continuously along vertical walls in standard configurations. Piping–seepage detectors, in such non-typical scenarios, are unable to establish a complete seepage flow field and can only identify infiltration areas, but cannot determine whether contact seepage or leakage channels exist behind walls. Considering the geometric constraints of underground spaces adjacent to levees and the detection blind zones at interfaces, this study adopts modified ground-penetrating radar and electrical resistivity tomography (ERT) for integrated detection in underground spaces adjacent to levees.

2.1.1. Ground-Penetrating Radar (GPR)

GPR is based on contrasts in the dielectric permittivity of subsurface media. High-frequency electromagnetic waves are emitted by a transmitting antenna (T) into the subsurface and are reflected at stratigraphic boundaries or targets, after which they are recorded by a receiving antenna (R). The levee body and foundation soils act as a complex filter, absorbing and reflecting electromagnetic waves to different degrees, so that the recorded radar signal integrates physical information from multiple subsurface media. By analyzing the two-way travel time, phase, waveform, and amplitude of reflected waves, the stratigraphic distribution, burial depth, and the geometry, scale, and position of defects can be determined.

For media like levee, soil and water, the wave velocity can be expressed as follows:

$$v={\displaystyle \frac{c}{\sqrt{{\epsilon }_r}}}$$

(1)

where c is the speed of light in vacuum, and ${\epsilon }_r$ is the relative dielectric permittivity.

The depth of a reflecting interface is calculated as follows:

$$D={\displaystyle \frac{v\cdot t}{2}}={\displaystyle \frac{c\cdot t}{2\sqrt{{\epsilon }_r}}}$$

(2)

where t is the two-way travel time of the radar wave.

A strong contrast in dielectric permittivity exists between concrete walls and water-bearing interface zones. When radar waves penetrate the concrete wall and encounter this interface, strong reflection signals are generated, enabling high-resolution imaging of hazard interfaces.

2.1.2. Electrical Resistivity Tomography (ERT)

ERT is based on contrasts in the electrical properties of rocks and soils. An electric current is injected into the subsurface through electrodes, and the distribution of conductive current under the applied electric field is measured to infer the occurrence and geometry of subsurface targets. As an array-based electrical method, ERT provides large data volumes per survey and high operational efficiency, and can simultaneously characterize resistivity variations in both horizontal and vertical directions.

In ERT surveys, current I is injected into the ground through electrodes A and B, and the potential difference ΔV is measured between electrodes M and N. The apparent resistivity at a given location is calculated as ρ_s = KΔV/I, where K is the geometric factor. The distribution of apparent resistivity is then used to interpret stratigraphy and identify anomalies within the survey area.

Leakage channels and water-rich contact seepage zones in levees typically exhibit low-resistivity characteristics, whereas uneroded concrete structures or dense soils show relatively high resistivity. This pronounced electrical contrast provides the physical basis for identifying hazards using ERT.

2.1.3. Coupling Mechanism of Seepage and Geophysical Responses

In urban levees with adjacent underground spaces, the significant difference in deformation characteristics between the rigid concrete retaining wall and the flexible levee soil results in a weak bonding at the soil-structure interface. This creates a vulnerable zone highly susceptible to soil particle migration, providing the geometric boundary conditions for contact seepage. Simultaneously, the natural hydraulic head difference between the high river water level and the adjacent underground space provides a strong hydrodynamic driving force for the seepage. Driven by this high head difference, the local hydraulic gradient and seepage force within poorly graded, high-permeability zones of the levee fill increase sharply. When the actual hydraulic gradient exceeds the critical hydraulic gradient of the soil, the flow continuously erodes and transports fine particles from the soil skeleton along the soil–wall interface. This process triggers internal erosion and ultimately evolves into continuous contact seepage zones and leakage channels.

The aforementioned hydraulic failure process directly leads to significant alterations in the macroscopic physical properties of the porous medium. According to the classical Archie’s Law, the bulk apparent resistivity of a porous medium is calculated as follows:

$${\rho }_t={\rho }_w\cdot a\cdot {\varphi }^{-m}\cdot S_w^{-n}$$

(3)

where ${\rho }_t$ is total formation resistivity, ${\rho }_w$ is pore water resistivity, $\varphi$ is porosity, $S_w$is water saturation, a is tortuosity, m is cementation, n is saturation exponents.

The internal erosion induced by contact seepage not only causes the loss of fine particles and a dramatic local increase in porosity, but the active seepage recharge also maintains the soil in a near-saturated state. These factors collectively result in a relatively low apparent resistivity in the affected areas. Therefore, the “extremely low-resistivity anomalies” captured by electrical resistivity tomography (ERT) behind urban levee walls are inherently not static geological heterogeneities in the traditional sense. Instead, they directly correspond, in a hydraulic context, to active leakage channels or contact seepage zones characterized by the most densely concentrated streamlines, the highest porosity, and maximum soil saturation.

2.2. Lateral Scanning Method

To overcome the limitations of conventional surface-based surveys in spatially constrained environments adjacent to levees, this study developed a detection scheme based on lateral scanning along vertical walls.

2.2.1. Lateral Scanning on Vertical Surfaces

Conventional surveys usually deploy survey lines horizontally on levee crests, adopting a downward-looking “half-space” detection mode. To focus on potential contact seepage hazards at the contact zone between underground spaces and levee soils, the data acquisition surface in this study is relocated to vertical walls within the underground space. Accordingly, the detection direction is transformed from vertical downward to horizontal lateral (toward the interior of the levee), so that current density lines or electromagnetic waves propagate as nearly perpendicular as possible to the hazard development surface. To obtain three-dimensional information on hazards behind the wall, multiple parallel survey lines are arranged on the wall at different elevations, allowing a three-dimensional characterization of hazard geometry. The combined ERT + GPR lateral scanning can achieve a typical survey speed of 100–150 m/h along the vertical wall. Compared with conventional surface-based geophysical methods that are often limited by inaccessible levee crests, hardened surfaces, and detection blind zones, the proposed method shows clear advantages in terms of adaptability to space-constrained urban environments and wall-coupling stability.

2.2.2. Electrode Coupling Technique on Hardened Surfaces

To address the problems that conventional steel spike electrodes cannot be driven into concrete walls and that contact resistance is excessively high, this study replaces point-type spike electrodes with patch-type electrodes which have a large contact area [15]. To meet the engineering requirements of easy on-site material acquisition and rapid deployment, this study utilized in situ cohesive soil mixed with industrial salt (NaCl) as a highly conductive coupling medium, and the electrodes are firmly fixed onto the vertical wall using strong adhesive tape (Figure 1). The long-term stability of the coupling agent was evaluated during field surveys. The moist clay–salt mixture remains sufficiently conductive for 2–4 h under typical urban indoor conditions. The key parameters of this coupling method and the quality control (QC) indicators for field data acquisition are summarized in

Table 1. The key parameters of this coupling method and the quality control (QC) indicators.

Category	Parameter	Specification
Electrode	material	pure copper patch
	dimensions	5 cm×5 cm
Coupling Medium	clay fraction	≥30%
	water content	40~50% approaching liquid limit
	application thickness	2~5 mm
	NaCl dosage	5~3% by mass of cohesive soil
	effective survey time window	2~4 h
QC Indicators	contact resistance	0.8~4.5 kΩ
	data rejection criteria	negative apparent resistivity, RMS error ≥ 5%, or I < 10 mA

. This technique effectively reduces the grounding resistance on dry concrete surfaces and ensures stable current injection for array measurements.

2.2.3. Ground-Penetrating Radar (GPR) Sidewall Detection Technique

When conventional handheld GPR antennas are used for sidewall surveys in underground spaces, practical limitations such as poor antenna–wall coupling, difficulties in precise positioning, and low operational efficiency are frequently encountered. To address these issues, a dedicated wall-coupled automatic GPR detection device for underground spaces was developed in this study. The system mainly consists of an omnidirectional mobile chassis, an adjustable lifting support frame, and an elastic wall-coupled GPR antenna module.

A radar sliding carriage structure incorporating an elastic buffering mechanism (compression springs) was specifically designed. By utilizing the restoring force of the springs to apply a constant normal pressure, the GPR antenna is maintained in continuous close contact with the wall surface during movement. This configuration effectively eliminates air-gap-induced reflection interference caused by antenna detachment and ensures stable electromagnetic coupling at the concrete interface, thereby improving signal quality and detection reliability.

2.2.4. Data Processing and Analysis

Given the pronounced non-standard characteristics of levee-adjacent detection environments, targeted processing of the acquired raw data is required to suppress environmental interference and enhance the interpretability of subsurface anomalies.

For ERT data, preprocessing and distortion point removal are first performed. To mitigate the shielding effect caused by high-resistance surface hardened layers—which typically leads to downward current deficiency and potential distortion—a ratio-based signal compensation algorithm based on the ratio method was employed.

$${\rho }_{\mathrm{cor}}={\rho }_{\mathrm{obs}}\cdot {\displaystyle \frac{{\rho }_{\mathrm{ref}}}{{\rho }_{\mathrm{surf}}+\lambda \cdot \max \left(0,{\rho }_{\mathrm{surf}}-{\rho }_{\mathrm{ref}}\right)}}$$

(4)

where ${\rho }_{cor}$ is the corrected apparent resistivity after signal compensation, ${\rho }_{obs}$ is the measured apparent resistivity of the vertical wall, ${\rho }_{surf}$ is the measured resistivity of the concrete wall (hardened surface), ${\rho }_{ref}$ is the background reference resistivity of the levee fill, and $\lambda$ is the regularization constraint coefficient, which is set to 0.2 in this study according to practical experience of geophysical detection.

The core principle of this algorithm involves utilizing the ratio of the background resistivity response of an unhardened site to that of the hardened layer as a correction factor. By multiplying this correction factor with the measured apparent resistivity obtained under hardened conditions, the algorithm effectively amplifies the masked subsurface geoelectrical information. This ratio-based strategy is specifically preferred over simple subtraction methods, as it inherently prevents the generation of non-physical negative resistivity artifacts in complex site conditions. Subsequently, geometric correction is conducted by mapping electrode coordinates arranged along the wall surface onto a two-dimensional inversion grid. Finally, nonlinear inversion is carried out using a damped least-squares approach. The high-resistivity shielding effect of the wall structure is explicitly incorporated into the inversion constraints, and the weighting of the initial model is optimized to improve the imaging resolution of low-resistivity anomalous bodies behind the wall.

For GPR data, zero-time correction is first implemented by accurately picking the onset time of the air–concrete interface to eliminate depth calculation errors. Background removal filtering is then applied to effectively suppress diffraction waves and horizontally propagating direct waves generated by the wall and surrounding environment, thereby enhancing reflections from the wall–back interface and deeper subsurface media boundaries. Finally, gain adjustment is performed to compensate for electromagnetic energy attenuation during penetration through high-resistivity wall materials and water-rich contact zones, resulting in enhanced amplitudes of deep reflection signals.

3. Results

To investigate the electromagnetic wave propagation behavior and electrical response characteristics associated with seepage defects along contact interfaces in levee-adjacent underground spaces, numerical simulations were conducted in Section 3 using gprMax (Version 3.1.6) and COMSOL Multiphysics (Version 5.6). Ground-penetrating radar (GPR) forward modeling and array resistivity method simulations were respectively performed for contact seepage scenarios, providing reference patterns for image interpretation and data analysis in practical field investigations.

gprMax is an electromagnetic numerical simulation software based on the finite-difference time-domain (FDTD) method, which solves Maxwell’s equations in three-dimensional space to simulate electromagnetic wave propagation processes. It has been widely applied in modeling and analysis of GPR responses.

COMSOL Multiphysics is a multiphysics coupling simulation platform based on numerical techniques such as the finite element method (FEM) and boundary element method. In the numerical simulation of the array resistivity method, the finite element method is primarily employed.

3.1. Construction of the Geological Conceptual Model

3.1.1. Geometric Model

The geophysical detection environment of levee-adjacent underground spaces exhibits pronounced non-standard characteristics. The core challenge lies in the fact that the target anomalies (potential hazards) are located behind a high-impedance concrete retaining wall. Moreover, these hazards typically manifest as horizontally extending, strip-shaped seepage channels, as well as thin vertical seepage zones developed along the contact interface.

Based on the vertical-planar-array lateral scanning detection approach described in the previous section, the levee-adjacent underground space is simplified in this study as a semi-infinite domain extending landward from the waterside face of the underground space wall. The geometric structure of the model mainly consists of a concrete wall, a contact seepage zone, seepage channels, and levee soil materials. The conceptual model is illustrated in Figure 2.

3.1.2. Model Parameters

In the gprMax simulations, a Ricker wavelet with good time–frequency localization was selected as the excitation source to represent the radar signal. A GPR antenna with a center frequency of 100 MHz was used. The spatial grid size was set to 0.01 m, the simulation time window was 5 × 10⁻⁷ s, and the acquisition step along the survey line was 0.1 m.

For the COMSOL Multiphysics simulations, a steady-state electric field formulation was adopted. A zero electric potential was assigned at infinity as the boundary condition. The computational domain was discretized using triangular finite elements, with the maximum element size not exceeding 0.02 m.

The model domain has a length of 10 m and a width of 5 m. The rectangular high-resistivity region at the top represents the concrete layer with a thickness of 0.5 m, corresponding to the exterior basement wall of levee-adjacent structures. This layer acts as the primary shielding medium impeding the penetration of electromagnetic waves and electrical current, with a resistivity of 10,000 Ω·m and a relative dielectric permittivity of 3.

An elongated elliptical zone with relatively lower resistivity represents the contact interface between the concrete wall and the levee soil, with a major axis length of 9 m and a minor axis length of 0.8 m. This zone simulates a sheet-like seepage band developed along the structural contact interface, characterized by a resistivity of 10 Ω·m and a relative dielectric permittivity of 20.

An inclined rectangular strip-shaped low-resistivity zone represents a concentrated seepage channel penetrating the levee body, with a width of 0.8 m, a resistivity of 1 Ω·m, and a relative dielectric permittivity of 30.

All remaining regions in the model are assumed to be homogeneous soil, with a resistivity of 20 Ω·m and a relative dielectric permittivity of 15. For all materials in the model, the relative magnetic permeability is set to 1, and magnetic losses are neglected. The geophysical parameters used in the model are summarized in

Table 2. Geophysical parameters of the levee-adjacent hazard model.

Material Type	Geometric Characteristics	Resistivity	Relative Permittivity	Geophysical Characteristics
Concrete layer	top rectangle	10,000	3	extremely high resistivity, low permittivity, dense hardened layer
Contact seepage	flat strip	10	20	low resistivity, medium-high permittivity, high water content
Leakage channel	inclined rectangle	1	30	extremely low resistivity, high permittivity, near-saturated
Background soil	/	20	15	medium-low resistivity, unsaturated soil

.

3.2. Analysis of Forward Modeling Results

3.2.1. Resistivity Response Characteristics

Figure 3 presents the forward modeling results of the levee-adjacent hazard model obtained using the Wenner array. The resulting resistivity section exhibits clear stratification extending from the concrete wall toward the levee body, and the spatial extent and general geometry of the seepage channel can be clearly identified.

The concrete wall layer (0–0.5 m) appears as a relative high-resistivity zone of 30–50 Ω·m. Although the true resistivity of concrete is set as 10,000 Ω·m in the model, the detected value is significantly lower, which is caused by the thin-wall averaging effect and current penetration. When the wall thickness is smaller than the electrode spacing, most injection current bypasses the high-resistivity concrete and enters the low-resistivity medium behind it, resulting in an apparent resistivity “dilution effect”.

The contact seepage zone presents a continuous layered low-resistivity anomaly of 10–20 Ω·m, which is 50–75% lower than the background soil (20 Ω·m). This significant contrast originates from the high water saturation and increased porosity caused by internal erosion, consistent with the variation law described by Archie’s law.

The concentrated seepage channel shows a closed low-resistivity anomaly of 5–10 Ω·m, which is 75–90% lower than the background soil. Its inclined shape and horizontal range (26–34 m) are completely consistent with the preset model, indicating that the lateral scanning ERT method can quantitatively locate and characterize the penetration seepage channel with high resolution.

The deep soil shows a relatively high resistivity of about 80 Ω·m, which is caused by the current convergence effect of the low-resistivity seepage zone. Most current flows preferentially through the seepage channel, resulting in low current density in the deep matrix and high apparent resistivity.

3.2.2. Radar Wavefield Characteristics

Figure 4 shows the simulated GPR reflection profile of the levee-adjacent hazard model obtained using a 100 MHz GPR antenna. The radar section is characterized by clearly identifiable, horizontally layered reflection events with continuous in-phase axes.

The first group of strong reflection interfaces corresponds to the boundary between the concrete wall and the contact-type seepage zone, exhibiting high-amplitude responses. This indicates that radar waves are particularly sensitive to layered structures. When voids or water-filled layers develop along the contact interface, a laterally continuous strong reflection band is formed.

No distinct reflections associated with the deep, inclined seepage channel are observed in the profile. This absence is attributed to two primary factors. First, significant electromagnetic energy attenuation occurs when high-frequency radar waves propagate through the shallow contact seepage zone, resulting in insufficient energy reaching greater depths. Second, the relatively large inclination angle of the seepage channel causes the incident radar waves to undergo predominantly non-specular reflections, making the effective reflected signals difficult for the surface receiving antenna to capture.

4. Case Study

4.1. Study Area Overview

The Qinhuai River levee runs through the central urban area of Nanjing. It is classified as a Grade I levee with a flood protection standard of a 100-year return period. The levee corridor is densely occupied by underground structures, including basements, sunken plazas, and culverts passing through the levee.

The study area is a typical levee-adjacent underground space section (Figure 5a). The levee is an earth embankment, with a recreational boat pier and sightseeing walkway located on the riverside. On the landside, a sunken commercial plaza is present, with the ground surface approximately 3 m lower than the levee crest elevation. During periods when the Qinhuai River water level exceeded the warning threshold, seepage occurred on the landside of the levee, with seepage outlets located approximately 0.5 m above the floor level of the sunken plaza.

Within the study area, the levee crest is occupied by commercial buildings and landscaped green belts. The contact interface between the levee and adjacent buildings consists of a vertical concrete retaining wall. Under such conditions, conventional geophysical survey methods face significant challenges, including difficulties in deploying continuous survey lines along the levee crest and the inability to install traditional steel-rod electrodes into the concrete wall surface.

To address these site-specific constraints, a lateral scanning survey scheme based on vertical wall surfaces was implemented in this study. To analyze seepage development at different elevations, two parallel horizontal survey lines were arranged on the levee-adjacent wall surface at a vertical spacing of 1 m (Figure 5b). During array resistivity data acquisition, patch-type electrodes were employed to reduce contact resistance. Based on on-site testing results, the Wenner array, which offers stronger resistance to environmental interference, was selected for data collection.

For GPR surveys, to achieve a balance between investigation depth and resolution, shielded dual-frequency antennas with center frequencies of 70 MHz and 300 MHz were used in a point-by-point acquisition mode. The GPR survey lines were collocated with the resistivity survey lines.

4.2. Detection Results and Interpretation

4.2.1. Analysis of Array Resistivity Survey Results

Figure 6 shows the apparent resistivity sections obtained along survey lines D1 (lower elevation line) and D2 (higher elevation line). The horizontal axis represents the distance along the wall surface, while the vertical axis represents the distance extending from the levee-adjacent wall toward the waterside.

The inversion results in Figure 6a indicate that the electrical structure of the medium behind the wall exhibits pronounced stratified zoning characteristics. Based on apparent resistivity contrasts, four typical zones can be identified. Within a depth of 0–0.5 m, a relatively high-resistivity zone is observed, with apparent resistivity values of approximately 300–500 Ω·m, corresponding to the concrete wall. At depths of 0.5–2.5 m, a relatively low-resistivity zone (Low-Res1) appears as a banded feature, with apparent resistivity values of approximately 50–100 Ω·m. Combined with the numerical simulation results discussed above, this zone is interpreted as a contact seepage zone with elevated water content.

In the region spanning 26–34 m along the horizontal axis and 2–9 m along the vertical axis, another relatively low-resistivity zone (Low-Res2) is identified, with apparent resistivity values of approximately 50–300 Ω·m. This zone exhibits weaker connectivity and a tendency to extend toward the waterside, suggesting a horizontally developed seepage region. All other areas below a depth of 2.5 m are characterized by high resistivity, generally exceeding 400 Ω·m, indicating well-compacted fill with relatively low moisture content.

The inversion results in Figure 6b are generally consistent with those in Figure 6a, confirming the reliability of the survey results. However, along survey line D2, the connectivity of Low-Res2 is more pronounced, indicating that a through-going seepage pathway has developed at this elevation. In addition, a localized low-resistivity anomaly (Low-Res3) is observed in the region between 18 and 24 m along the horizontal axis and below a depth of 3 m. As this anomaly is not connected to Low-Res1, it is interpreted as a localized water-bearing zone.

Overall, the field array resistivity results are in good agreement with the numerical simulation outcomes. The resistivity sections display clear stratification: the concrete layer appears as a relatively high-resistivity, layered zone; the contact seepage zone is expressed as a band-shaped low-resistivity anomaly with finite thickness; seepage channels manifest as low-resistivity bands extending toward the waterside; and the levee fill is characterized by relatively high resistivity. Compared with the lower survey line (D1), the higher survey line (D2) exhibits better connectivity of seepage channels and the presence of localized water-bearing bodies, indicating more severe seepage development at higher elevations and a tendency for downward progression.

4.2.2. Analysis of Ground-Penetrating Radar Survey Results

Figure 7 presents the GPR results obtained along survey lines D1 (lower elevation line) and D2 (higher elevation line). As shown in Figure 7a, the high-frequency GPR data along line D1 exhibit clear shallow-layer stratification. A distinct interface is observed at a depth of approximately 2 m, which corresponds well with the position of the relatively low-resistivity zone (Low-Res1) identified by the array resistivity method. In contrast, as shown in Figure 7b, the low-frequency GPR data along line D2 do not reveal a well-defined seepage channel geometry, instead displaying disordered and weak diffracted wave patterns.

Overall, the field GPR results are in good agreement with the numerical simulation outcomes. Strong-amplitude zones and clear interfaces are observed in the shallow part of the profiles, whereas waveforms at greater depths are chaotic, making it difficult to identify the geometry and location of seepage channels. This limitation is attributed to the presence of a water-rich, low-resistivity layer in the shallow contact zone (0.5–2.5 m), which strongly attenuates radar wave energy and restricts the imaging capability of GPR for deep, steeply inclined seepage channels below 3 m.

4.3. Integrated Interpretation and Validation

By integrating the results of the array resistivity method and the GPR surveys, it can be inferred that a low-resistivity, water-bearing zone is widely developed within approximately 2 m from the wall extending toward the waterside, indicating the presence of contact seepage at the junction between the underground space and the levee soil. A through-going seepage channel is identified within the horizontal distance range of 28–34 m, with more severe seepage development observed along the higher survey line (D2), suggesting a tendency for downward progression. In addition, a closed low-resistivity anomaly within the range of 18–24 m is interpreted as a localized water-bearing body.

Seepage had already been observed during the underground space investigation at this site. Field observations indicate that the seepage outlet elevation lies between survey lines D1 and D2. Borehole investigations further reveal that the levee body within the horizontal distance range of 28–34 m is predominantly composed of plain fill and miscellaneous fill. The plain fill is dominated by silty clay with moderate permeability (approximately 7 × 10⁻³ cm/s), whereas the miscellaneous fill is rich in gravel, concrete fragments, and coarse sand, exhibiting high permeability (approximately 3 × 10⁻² cm/s). The strong reflections observed in GPR images result from the significant dielectric contrast between the concrete wall and the water-saturated miscellaneous fill, which matches the low-resistivity characteristics of the contact seepage zone revealed by ERT.

5. Conclusions

(1)

The vertical-array lateral scanning detection method proposed in this study effectively overcomes the spatial constraints of “no accessible ground surface for surveying” commonly encountered in urban levee-adjacent environments. By introducing patch-type electrodes in combination with high-conductivity coupling agents, the technical challenges associated with current injection and excessive contact resistance on hardened concrete wall surfaces were successfully resolved. Stable current injection and high signal-to-noise ratio data acquisition for the array resistivity method on vertical hardened surfaces were achieved. Field applications demonstrate that this method can penetrate high-resistivity concrete walls and obtain clear electrical resistivity images of the soil structure behind the wall, confirming the engineering feasibility of shifting the observation surface from a horizontal ground plane to a vertical wall surface.

(2)

Comparative analysis of numerical simulations and field measurements indicates that concrete walls are expressed as relatively high-resistivity zones rather than extremely high-resistivity bands. Contact-type seepage is characterized by a layered low-resistivity zone immediately behind the wall, while deep seepage channels manifest as vertically continuous low-resistivity anomalies extending toward the waterside.

(3)

The results from the ground-penetrating radar and array resistivity surveys demonstrate the inherent limitations of using a single geophysical method in levee-adjacent environments. As GPR cannot effectively image deep seepage channels because of strong signal attenuation in shallow water-rich and high-conductivity zones behind the wall, the effective imaging depth of GPR is mainly limited to 2–3 m in this levee site condition. Below the depth of 3 m, the electromagnetic signal amplitude decays significantly, and the reflected wave energy is too weak to distinguish the spatial morphology of deep seepage channels accurately. This limitation should be fully considered in practical applications. By contrast, ERT remains sensitive to deep seepage pathways and compensates for the GPR blind zone, highlighting the necessity of the combined ERT–GPR strategy for reliable hazard identification.

(4)

Despite its demonstrated effectiveness, the proposed vertical-array lateral scanning method has certain applicability boundaries and technical limitations. The efficacy of GPR may be significantly reduced in highly conductive environments (such as highly saturated clay or saline soils) due to severe electromagnetic wave attenuation, or when facing strong signal scattering caused by dense steel rebar networks within the concrete retaining walls. Meanwhile, ERT inversion inherently exhibits non-uniqueness and ambiguity, particularly when delineating complex three-dimensional seepage geometries using two-dimensional profiles, or when operating under strong urban stray current interference. Therefore, a comprehensive hazard assessment should always integrate these complementary geophysical methods with local hydrogeological data and physical borehole validations.

In summary, the integrated geophysical detection approach based on vertical-surface lateral scanning proposed in this study provides a non-destructive, efficient, and quantitative technical solution for hazard investigation in levee-adjacent and levee-crossing engineering projects under complex urban conditions. This methodology offers valuable methodological support for the qualitative identification of levee-related hidden hazards in environments characterized by strong urban interference.