An Integrated Approach for Earth Infrastructure Monitoring Using UAV and ERI: A Systematic Review

Abstract

The integrity of earth infrastructure, encompassing slopes, dams, pavements, and embankments, is fundamental to the functioning of transportation networks, energy systems, and urban development. However, these infrastructures are increasingly threatened by a range of natural and anthropogenic factors. Conventional monitoring techniques, including inclinometers and handheld instruments, often exhibit limitations in spatial coverage and operational efficiency, rendering them insufficient for comprehensive evaluation. In response, Uncrewed Aerial Vehicles (UAVs) and Electrical Resistivity Imaging (ERI) have emerged as pivotal technological advancements, offering high-resolution surface characterization and critical subsurface diagnostics, respectively. UAVs facilitate the detection of deformations and geomorphological dynamics, while ERI is instrumental in identifying zones of water saturation and geological structures, detecting groundwater, characterizing vadose zone hydrology, and assessing subsurface soil and rock properties and potential slip surfaces, among others. The integration of these technologies enables multidimensional monitoring capabilities, enhancing the ability to predict and mitigate infrastructure instabilities. This article focuses on recent advancements in the integration of UAVs and ERI through data fusion frameworks, which synthesize surface and subsurface data to support proactive monitoring and predictive analytics. Drawing on a synthesis of contemporary research, this study underscores the potential of these integrative approaches to advance early-warning systems and risk mitigation strategies for critical infrastructure. Furthermore, it identifies existing research gaps and proposes future directions for the development of robust, integrated monitoring methodologies.

1. Introduction

Earth infrastructure, including slopes, dams, pavements, and embankments, forms the backbone of transportation, energy, and urban development systems. These structures, characterized by their interaction with soil, rock, and water, are critical for societal functionality and economic growth. Their role in transportation is particularly vital, as they support highways, railways, and bridges, ensuring connectivity and trade. However, earth infrastructure is constantly exposed to natural and anthropogenic challenges, making effective monitoring essential to prevent failures that could disrupt operations [1,2] or endanger lives [3,4].

Traditionally, monitoring of slopes and embankments has relied on methods such as inclinometers, total stations, piezometers, extensometers, and handheld GPS devices [5]. These tools provide localized measurements of displacement, tilt, water pressure, and other parameters [6]. Although effective in controlled environments, they are often labor-intensive, time-consuming, and limited in coverage, making them less practical for monitoring large or inaccessible areas. In addition, reliance on manual data collection and interpretation can introduce errors and delays, reducing the timeliness and precision of risk assessments. These limitations underscore the need for innovative approaches that offer greater spatial coverage, higher temporal resolution, and enhanced safety.

Uncrewed Aerial Vehicles (UAVs) and Electrical Resistivity Imaging (ERI) have emerged as transformative technologies for earth infrastructure monitoring. UAVs address the limitations of traditional methods by providing high-resolution, real-time surface data, enabling early detection of surface deformations and geomorphological changes [7,8,9,10]. ERI complements this by offering insights into subsurface conditions, such as detecting perched water conditions, characterizing unsaturated zones, and identifying potential failure surfaces, which are crucial for understanding and mitigating failure risks. Together, these technologies provide a multidimensional perspective, enhancing the ability to anticipate and manage instability in earth structures.

UAV sensor technologies [11], including RGB cameras, LiDAR, infrared imaging, multispectral imaging, and InSAR [12], bring versatility to earth infrastructure monitoring. RGB imaging offers high-resolution visual data for detecting surface anomalies, while LiDAR excels in creating detailed topographical models, even in vegetated areas. Infrared imaging detects temperature variations, which are indicative of moisture content and seepage, and multispectral imaging captures data across multiple wavelength bands, enabling advanced vegetation analysis and detection of soil and material composition. InSAR monitors subtle surface displacements over time [13]. When integrated into UAV systems, these sensors enable comprehensive surface monitoring across diverse environmental and geological conditions.

Electrical Resistivity Imaging (ERI) complements UAV capabilities by providing critical insights into subsurface conditions [14,15]. This geophysical technique measures variations in soil and rock resistivity, allowing for moisture content profiling [16], the detection of saturated zones, the monitoring of groundwater conditions [17], and the identification of fractures [18], karst fracture zones [19], or slip surfaces. Advanced ERI methodologies, such as time-lapse imaging and integration with Ground Penetrating Radar (GPR) [20,21], enable dynamic monitoring and deeper subsurface investigations. These capabilities are particularly valuable for understanding and mitigating failure risks associated with expansive soils, perched water tables, and other geotechnical challenges.

This article systematically analyzes UAV and ERI technologies, emphasizing their integration through the development and application of data fusion frameworks. By synthesizing findings from recent studies, this paper highlights how combining UAV-derived surface data with ERI-based subsurface characterization can deliver comprehensive assessments of earth infrastructure. These data fusion frameworks enable the seamless integration of multidimensional datasets, facilitating enhanced visualization, real-time monitoring, and predictive analytics. Such advancements are critical for proactive monitoring and early warning systems [22,23], which address the complex challenges associated with the stability of slopes, dams, pavements, and other earthworks. By focusing on the synergy between UAV and ERI, we aim to demonstrate the potential of integrated approaches in delivering actionable insights for improved inspections, maintenance, and risk mitigation strategies.

The remainder of this paper is organized as follows: Section 2 introduces UAVs and their sensor technologies for surface monitoring, while Section 3 delves into subsurface monitoring using ERI. Section 4 discusses data fusion frameworks and their role in integrating surface and subsurface insights. Section 5 identifies the current research gaps and future directions.

2. Surface Monitoring Using UAVs

Efficient monitoring of highway embankments and slopes is critical for detecting early warning signs of instability and preventing structural failures [24]. UAVs have revolutionized surface monitoring by enabling high-resolution, real-time data collection across vast and otherwise inaccessible areas [25,26]. By employing advanced imaging technologies, UAVs provide detailed surface models and visualizations essential for identifying deformation patterns, geomorphological changes, and other critical indicators of instability. The capabilities of UAV-based monitoring extend across multiple imaging modalities, including RGB (Red–Green–Blue), thermal, LiDAR (Light Detection and Ranging), multispectral, and InSAR (Interferometric Synthetic Aperture Radar) technologies, each contributing uniquely to comprehensive slope monitoring and risk mitigation, as shown in Figure 1. The remainder of this section discusses each of these imaging technology separately.

2.1. RGB Imaging

High-resolution RGB cameras mounted on UAVs are instrumental in capturing detailed surface imagery [28]. Through photogrammetry techniques such as Structure from Motion (SfM) [29], RGB data can be transformed into Digital Elevation Models (DEMs) and ortho-rectified images [11,30,31]. Photogrammetry for RGB cameras is a remote sensing technique that extracts precise geometric and spatial information from overlapping images to create high-resolution 3D models and orthomosaics. In Structural Health Monitoring (SHM), RGB photogrammetry aids in detecting surface deformations, cracks, and structural instabilities by providing detailed visual and geometric data of infrastructure [29]. These models enable precise visualization and measurement of surface features, such as cracks, landslide scars, and erosion patterns [32]. This involves solving for camera positions and scene geometry simultaneously, minimizing reprojection errors through bundle adjustment, which can be formally represented as follows:

$$\underset{\mathrm{P},\mathrm{X}}{min}\sum _{i,j}{\parallel {\mathrm{x}}_{ij}-\pi ({\mathrm{P}}_i,{\mathrm{X}}_j)\parallel }^2$$

(1)

where:

• ${\mathrm{P}}_i$: Camera parameters for image i;
• ${\mathrm{X}}_j$: 3D point j;
• ${\mathrm{x}}_{ij}$: Observed 2D projection of point j in image i;
• $\pi$: Projection function.

This optimization refines both camera parameters and 3D point positions to achieve the best fit between the projected and observed image points, enhancing the accuracy of the reconstructed models [29].

The precision of these reconstructions and analyses relies heavily on factors such as image overlap, camera calibration, and Ground Control Point (GCP) distribution. Deliry et al. [33] demonstrated that strategic flight path planning and optimized GCP placement can significantly enhance the accuracy of UAV-derived DEMs.

Nobahar et al. [1] utilized UAV-mounted RGB cameras to monitor embankment surfaces, creating high-resolution DEMs to detect surface deformations, cracks, and displaced soil volumes. The accuracy of the 3D models enabled the identification of early signs of distress in expansive clay soils, facilitating proactive maintenance measures. Similarly, the authors in [34] applied RGB imaging on highway slopes in Mississippi following significant rainfall events. The high spatial resolution of the images allowed precise measurement of surface displacement, leading to targeted interventions to mitigate failure risks.

On the other hand, Hao et al. [28] explored an expanded application of RGB imagery by combining it with multispectral data for vegetation analysis on highway slopes. Using a DJI Phantom 4 UAV, the authors captured RGB images to distinguish between bare soil and vegetated areas, further deriving vegetation indices such as NDVI (Normalized Difference Vegetation Index) and ARVI. These indices provided insights into vegetation health, chlorophyll content, and biomass density—key indicators for slope stability, particularly in erosion-prone regions.

In more complex terrain, Solla et al. [35] integrated UAV-based RGB imaging with Ground Penetrating Radar (GPR) for infrastructure monitoring. This dual approach enabled detailed surface mapping in challenging environments, such as retaining walls and tunnels, where ground-based methods are infeasible. The RGB imagery enhanced visual documentation, while GPR provided critical subsurface insights.

Shang et al. [36] further extended the application of UAV-mounted RGB cameras by combining optical imagery with Infrared (IR) imaging for bridge inspections. The RGB data captured visible surface deformations, such as cracks and defects. At the same time, IR imaging highlighted temperature variations indicative of moisture accumulation or material delamination, providing a more comprehensive structural health assessment.

UAVs equipped with RGB cameras are widely used for structural health monitoring, particularly for detecting surface defects such as cracks on bridge decks, runways, and concrete pavements. High-resolution aerial imagery captured by UAVs provides a cost-effective and efficient method for identifying early signs of structural deterioration. By employing photogrammetry and image processing techniques, UAV-collected RGB data can be analyzed to detect and classify surface defects, allowing for timely maintenance interventions. In a study by [37], high-resolution RGB aerial images were collected using a DJI Phantom 3 UAV, following a structured flight plan to ensure comprehensive coverage of the study area. The UAV operated at multiple altitudes (40 m, 50 m, 60 m) with a 75% front and side overlap to facilitate seamless image stitching and orthomosaic generation. The images were captured under consistent lighting conditions to minimize shadow effects and enhance image clarity. The collected RGB data were processed using photogrammetry techniques to generate georeferenced orthomosaics, which were later analyzed for structural and vegetation assessments. Deep learning-based crack detection models further enhance accuracy by automating the identification of damage patterns, contributing to improved infrastructure management and safety.

These studies [1,11,28,29,30,31,34,35,36,37] collectively showcase the versatility of UAV-based RGB imaging for slope and infrastructure monitoring, highlighting its role in detecting surface anomalies, mapping deformation zones, and supporting integrated monitoring frameworks.

2.2. Thermal Imaging

Thermal cameras mounted on UAVs enhance surface monitoring by detecting temperature variations linked to subsurface moisture and seepage [36]. These variations are governed by the Stefan–Boltzmann Law [38], which is formally represented as follows:

$$E=\sigma T^4$$

(2)

where:

• E: Emitted thermal radiation (W/${\mathrm{m}}^2$);
• $\sigma$: Stefan–Boltzmann constant ($5.67\times {10}^{-8}$ W/${\mathrm{m}}^2$/${\mathrm{K}}^4$);
• T: Surface temperature (K).

The principle underlying thermal imaging is that regions with higher moisture content tend to exhibit cooler temperatures due to evaporative cooling, whereas dry zones or areas with subsurface issues (e.g., air gaps, fractures) retain heat and appear warmer [39]. This thermal contrast enables the identification of areas prone to failure caused by moisture retention, seepage, or soil deformation.

Thermal imaging UAVs [40] collect data using thermal infrared sensors, typically operating within the 8–14 µm spectral range. To ensure comprehensive thermal mapping, UAVs follow a pre-planned flight path at optimized altitudes and overlap settings. Primary data sources include radiometric thermal images captured under controlled environmental conditions, while secondary sources, such as meteorological data, aid in calibration and atmospheric correction. Ground reference panels with known emissivity values are often used for temperature calibration, along with infrared thermometers for validation. Data collection is usually conducted under stable weather conditions to minimize atmospheric interference, ensuring precise surface temperature measurements for applications such as vegetation health assessment, structural inspections, and environmental monitoring.

Salunke et al. [34] applied UAV-mounted thermal imaging to assess near-surface moisture levels on highway embankments following heavy rainfall events. By capturing diurnal variations in Land Surface Temperature (LST), the authors developed a predictive model correlating cooler zones with moisture retention, which posed risks of slope instability. The study demonstrated how UAV thermal imagery could detect early signs of water-driven failures and enable timely interventions.

Similarly, Nobahar et al. [1] used thermal imaging to analyze surface temperature distributions on embankments. The study showed that areas with high moisture content consistently exhibited lower temperatures due to evaporative cooling, while dry zones had relatively higher thermal signatures. This approach provided a non-invasive method for identifying moisture accumulation zones that triggered slope movement, particularly in expansive clay soils.

Shang et al. [36] extended the use of UAV-mounted thermal cameras beyond slope monitoring, integrating optical (RGB) and thermal imaging for bridge deck assessments. The thermal imagery successfully identified temperature anomalies indicative of subsurface moisture infiltration or material delamination, which are precursors to structural degradation. This combined approach demonstrated the utility of thermal imaging in both slope and structural health monitoring.

In contrast to RGB imaging, thermal data can penetrate environmental factors such as low light conditions or slight vegetation cover, enabling continuous monitoring regardless of the time of day. Studies emphasize the importance of capturing diurnal thermal variations—temperature differences between day and night cycles—as these variations improve the reliability of moisture detection and thermal anomaly mapping [1,34].

Thermal imaging techniques are increasingly applied to various functionalities such as (i) identifying seepage pathways and zones with subsurface moisture accumulation, (ii) detecting anomalies in pavement surfaces due to water infiltration, and mapping thermal gradients that correspond to soil moisture variations or weak zones.

These capabilities make UAV-based thermal imaging particularly effective in identifying moisture-driven failure risks, such as those encountered in rainfall-induced slope instabilities or embankment distress [16]. When integrated with other monitoring technologies like ERI or RGB imaging, thermal imaging offers complementary insights, improving the overall accuracy and efficiency of slope monitoring frameworks.

2.3. LiDAR

LiDAR technology offers unmatched precision in capturing elevation data, even through dense vegetation [31]. Distances are measured by emitting laser pulses and recording the time it takes for reflections to return, enabling the calculation of distances, which can be formally represented as follows:

$$d={\displaystyle \frac{c·\Delta t}{2}}$$

(3)

where:

• c: Speed of light (3 × ${10}^8$ m/s);
• $\Delta t$: Time interval between pulse emission and reception.

This formula enables precise distance measurements by calculating the time it takes for a laser pulse to travel to an object and back [41]. The emitted pulses generate dense 3D point clouds, which can be processed into Digital Terrain Models (DTMs) and Digital Surface Models (DSMs), offering high-resolution representations of slope surfaces. This capability makes LiDAR invaluable for analyzing topographical changes, quantifying volumetric displacements, and identifying subtle deformation patterns in landslide-prone regions [42].

LiDAR-equipped UAVs collect data by emitting laser pulses toward the ground and measuring the time taken for the reflected signals to return to the sensor. The system consists of a LiDAR scanner, an onboard GNSS receiver for georeferencing, and an Inertial Measurement Unit (IMU) for tracking orientation and movement. UAVs follow a predefined flight path with optimized altitude and overlap settings to ensure uniform data coverage. The collected point cloud data, often georeferenced in a global coordinate system, are processed to generate high-resolution digital elevation models (DEMs), terrain models, and 3D structural reconstructions. LiDAR UAVs are widely used for topographic mapping, vegetation analysis, infrastructure inspection, and disaster management applications due to their high accuracy and efficiency in capturing detailed terrain features [43].

Salunke et al. [34] employed UAV-mounted LiDAR systems to produce georeferenced 3D models of highway slopes following rainfall events. The study demonstrated how LiDAR data facilitated the precise mapping of surface deformations and subsurface moisture zones, providing insights into slope instability mechanisms.

In more specialized applications, Sreenath et al. [44] showcased the capability of UAV-mounted LiDAR in detecting subtle deformations over time, particularly in embankments and large civil structures. This study emphasized the value of LiDAR systems for tracking long-term changes, which are critical for early failure detection and slope stability monitoring.

Vishweshwaran et al. [31] highlighted the versatility of UAV LiDAR systems beyond slope monitoring. The authors demonstrated applications for pavement condition assessment, including detecting cracks, rutting, and surface distress, as well as underground infrastructure monitoring in tunnels and shafts where human access is restricted. The ability of LiDAR to capture detailed geometric features under complex conditions makes it an indispensable tool for comprehensive infrastructure health assessments.

Compared to traditional imaging techniques, like RGB, LiDAR’s penetration through vegetation enables accurate elevation mapping even in heavily vegetated or obstructed areas [45]. Additionally, LiDAR data facilitate volumetric analysis of landslide debris and soil displacements, which are crucial metrics for slope stability analysis and the design of remedial measures [46].

2.4. Multispectral Imaging

Multispectral cameras mounted on UAVs enable data capture across multiple spectral bands, such as visible, near-infrared, and red-edge, providing critical insights into surface vegetation health and other spectral attributes [47]. These cameras are handy for analyzing slope stability indirectly by evaluating vegetation cover and its relationship to erosion and instability risks [48].

Using multispectral-generated indices, including NDVI (Normalized Difference Vegetation Index), ARVI (Atmospherically Resistant Vegetation Index), and NDRE (Normalized Difference Red Edge), can provide indicators of vegetation health, chlorophyll content, and biomass density [49]. Healthy vegetation stabilizes slopes by mitigating soil erosion, while sparse vegetation can indicate areas vulnerable to erosion and slope failure [48].

Multispectral UAVs collect data using specialized sensors that capture reflectance across multiple discrete spectral bands, typically in the visible, Near-Infrared (NIR), and Shortwave Infrared (SWIR) regions. These UAVs follow a pre-programmed flight path with optimized altitude, overlap, and sensor settings to ensure comprehensive coverage of the target area. The collected multispectral images are georeferenced using onboard GPS and IMU systems, allowing for accurate spatial alignment. Pre-processing steps such as radiometric and geometric corrections are applied to enhance data accuracy. The processed multispectral imagery is then used for applications such as vegetation health monitoring, land cover classification, and soil moisture estimation by analyzing spectral indices like NDVI and NDRE [50].

Hao et al. [28] employed multispectral imaging using a DJI Phantom 4 Multispectral UAV to create vegetation probability maps that correlated vegetation health with slope stability. Similarly, Salunke et al. [34] used multispectral cameras to monitor vegetation patterns in regions prone to rainfall-induced slope instability, revealing areas with sparse vegetation at higher risk of failure.

Multispectral imaging has several advantages, including (i) detecting subtle vegetation stress before visible damage occurs, (ii) mapping erosion-prone areas based on sparse vegetation patterns, and (iii) complementing LiDAR and RGB imaging to improve slope stability models.

2.5. InSAR

InSAR is a remote sensing technique that leverages radar signals to detect ground displacement with millimeter-scale accuracy [51,52]. By measuring phase differences between radar acquisitions taken at different times, InSAR can quantify ground movement [32]. The displacement $d_{\mathrm{LOS}}$ along the line of sight for measuring surface deformations [53] can be calculated as follows:

$$d_{\mathrm{LOS}}={\displaystyle \frac{\lambda \Delta \varphi }{4\pi }}$$

(4)

where:

• $\lambda$: Radar wavelength;
• $\Delta \varphi$: Phase difference between acquisitions.

Compared to traditional satellite-based imaging [54], UAV-mounted InSAR systems are emerging as effective alternatives for localized, high-resolution monitoring [32]. The advantage of UAV-mounted systems lies in their ability to operate at lower altitudes, providing finer spatial resolution and greater flexibility for monitoring smaller, high-risk areas such as highway embankments and unstable slopes [55].

Interferometric Synthetic Aperture Radar (InSAR) UAVs collect data by emitting microwave radar signals towards the Earth’s surface and measuring the phase differences between multiple passes over the same area. These UAVs follow a predefined flight path at a consistent altitude to ensure accurate interferometric analysis. The collected radar data are georeferenced using onboard GPS and IMU systems to track position and orientation. Interferograms are generated by comparing phase differences from successive radar images, allowing for the detection of millimeter-scale ground deformation. The processed data are used for applications such as land subsidence monitoring, infrastructure stability assessment, and detecting surface deformation caused by landslides or mining activities [56].

Jiao et al. [32] demonstrated the effectiveness of InSAR by integrating UAV-derived optical imagery with Sentinel-1 InSAR data to monitor slope deformations in landslide-prone areas. Their study utilized Sentinel-1 data from 2017 to 2019 to create a time series of surface deformation measurements, revealing deformation signals as low as 5 mm/year up to two years before a significant landslide event. By correlating these subtle surface movements with UAV imagery, the authors identified surface anomalies that strongly correlated with subsurface water movement, providing critical insights into slope instability mechanisms. This comprehensive framework highlighted the potential of InSAR for detecting early warning indicators of slope failures, enabling proactive risk mitigation strategies.

UAV-mounted InSAR systems have also been applied to localized monitoring scenarios requiring higher temporal and spatial resolutions. For instance, combining UAV InSAR data with Digital Elevation Models (DEMs) enhances the interpretation of deformation patterns in complex terrains [57]. This integration provides a holistic understanding of surface dynamics, improving slope failure prediction accuracy.

Compared to other remote sensing technologies like LiDAR [58] or optical photogrammetry, InSAR excels in (i) Long-term monitoring: detecting cumulative deformations over extended periods. (ii) High sensitivity: capturing subtle displacements with millimeter precision. (iii) All-weather operation: operating effectively under cloud cover or low-visibility conditions.

However, vegetation interference and phase decorrelation remain significant challenges for InSAR-based technologies. Therefore, integrating InSAR with UAV-derived RGB and LiDAR [59] data is recommended to mitigate these issues, offering a more robust slope monitoring framework.

2.6. Comparison of UAV Imaging Sensors

The capabilities and applications of the imaging sensors discussed in this section are summarized in

Table 1. Comparison of UAV imaging sensors.

Sensor Type	Capabilities	Applications	Limitations
RGB Cameras	Captures high-resolution visible spectrum images; useful for creating DEMs and identifying surface features.	Surface deformation mapping, crack detection on bridges and pavements, monitoring ground subsidence, and vegetation analysis (e.g., NDVI) for infrastructure assessment [37].	Limited to daylight conditions; less effective in dense vegetation or occluded areas.
Thermal Cameras	Detects temperature variations to infer subsurface moisture and seepage.	Detecting seepage pathways in dams, identifying moisture retention zones in agriculture, and mapping soil anomalies for construction and land management [40].	Limited spatial resolution; cannot directly detect geological features; affected by environmental heat sources.
LiDAR	Generates precise 3D point clouds and terrain models; penetrates vegetation for accurate elevation mapping.	Topographical analysis, terrain mapping, landslide debris assessment, and infrastructure deformation monitoring.	Affected by atmospheric conditions; requires higher processing and equipment costs.
InSAR	Captures ground displacement with millimeter precision; effective in all-weather conditions.	Long-term deformation monitoring, subsidence detection, and identifying early signs of slope movement.	Vegetation interference and phase decorrelation may affect accuracy; lower spatial resolution compared to LiDAR.
Multispectral Cameras	Captures data across multiple spectral bands, enabling analysis of vegetation health and soil conditions.	Vegetation classification, erosion detection, and slope stability monitoring (e.g., NDVI, ARVI).	Limited spatial resolution compared to RGB and LiDAR; high sensitivity to lighting conditions.

, which highlights the complementary strengths of UAV-based imaging technologies. For example, RGB and thermal cameras excel in surface-level detail, while LiDAR and InSAR provide subsurface and long-term monitoring capabilities [60]. Multispectral imaging bridges the gap by offering additional spectral insights, particularly for vegetation and soil analysis. Integrating these sensors enables a comprehensive understanding of slope stability, particularly in challenging or high-risk environments. UAV imaging technologies offer diverse capabilities for monitoring geotechnical infrastructure, with each sensor type contributing uniquely to structural health assessments.

2.7. UAV Applications in Infrastructure Monitoring

Extensive research and practical implementations have demonstrated UAV imaging’s effectiveness in various infrastructure applications beyond embankments and slopes. Notable applications include the following:

• Bridge Decks and Asphalt Pavements: UAV-mounted RGB and thermal cameras have been widely utilized to detect surface cracks, rutting, and pavement distress in bridge decks and asphalt-covered roadways. High-resolution photogrammetry allows for detailed mapping of distress patterns, improving early detection and maintenance strategies [36].
• Airport Runways and Taxiways: UAV-based crack detection has proven effective in assessing the condition of concrete and asphalt surfaces in high-traffic areas such as airport runways and taxiways. The combination of RGB imaging and infrared thermography enables the identification of surface wear, thermal expansion effects, and structural anomalies [61].
• Port and Airport Concrete Pavements: UAV imaging, particularly LiDAR and multispectral sensors, has been applied for assessing large-scale concrete pavements in ports and airports. These technologies assist in detecting joint displacement, subsurface voids, and moisture infiltration that could compromise pavement integrity [62].
• Rubble Mound Structures and Change Detection: UAV-based LiDAR and InSAR technologies have been instrumental in monitoring instability and morphological changes in rubble mound breakwaters, dams, and embankments. These sensors capture deformation patterns and settlement issues over time, supporting predictive maintenance and risk mitigation strategies [55].

3. Subsurface Monitoring Using ERI Technology

Understanding subsurface conditions is pivotal in assessing the stability of slopes and embankments, particularly in areas prone to climate-induced failures, problematic soil or expansive soils, or underlying geological weaknesses [63]. Electrical Resistivity Imaging (ERI) has been proven to be a transformative geophysical technique for non-invasive subsurface monitoring. By measuring variations in electrical resistivity, ERI identifies critical subsurface features, such as saturated and unsaturated zones, fractures, and potential slip surfaces [64]. These insights are instrumental in complementing surface data from UAVs and enhancing the overall stability analysis.

3.1. Theory of Electrical Resistivity

ERI is a geophysical method that involves the use of two transmitting electrodes to inject current, thereby generating an electric potential field. The subsequent changes in potential are then measured between two receiving electrodes [65]. Electrical resistivity measures the degree to which a material resists the passage of electric current. The distance between the transmitting and receiving electrodes influences the spatial resolution and the volume of interrogation for each measurement collected. The ERI theory is relevant in a completely homogeneous half-space medium. The calculation of soil resistivity values for the subsurface involves determining the magnitude of the injected current and measuring the resultant electric potential at designated locations. Subsurface homogeneity is infrequently observed in natural conditions. When an electric current is injected, it will traverse the path of least resistance [65].

The electrical resistivity of soil follows Ohm’s law, where potential differences are quantified by the introduction of artificially generated currents into the soil [66]. For an electric field applied in a resistivity survey, Ohm’s law provides the foundation for calculating resistivity ($\rho$), defined as:

$$\rho ={\displaystyle \frac{\Delta V}{I}}·K$$

(5)

where:

• $\Delta V$: Voltage difference;
• I: Current;
• K: Geometric factor dependent on electrode spacing and configuration.

The geometric factor K is crucial as it accounts for the geometry of the electrode array used during the survey, influencing the depth and resolution of the investigation [67].

In a resistivity survey, apparent resistivity ${\rho }_a$ is typically measured, as soil is rarely homogeneous. The effective resistivity can then be calculated by inversion processes to reflect the actual subsurface distribution. Two methods exist for processing and interpreting electrical resistivity data: (1) inversion and (2) forward modeling. Completion of one process is contingent upon the execution of the other process. An image is provided that illustrates the conversion of field investigation data, followed by the calculation of apparent resistivity based on the injected current during the test and the resulting potential difference. A pseudosection is generated using the defined array type and the observed apparent resistivity values. The pseudosection supplies the essential data required for inversion.

The electrical resistivity is affected by various factors, including sub-surface soil moisture content and temperature [68], degree of saturation [69], organic content [70], pore water composition [71], geologic formation [72], ion concentration in pore water, as well as soil texture and structure [73,74]. Soil resistivity is influenced by several factors, with soil moisture and matric suction being particularly significant. These factors have a direct impact on the electrical conductivity of the soil and the distribution of pore water. Soil moisture has a substantial impact on soil resistivity, as the presence of water in the soil increases ionic mobility, leading to a decrease in resistivity [75]. Research indicates that resistivity exhibits an exponential decrease as moisture content increases, especially in fine-grained soils where pore water plays a significant role in the conduction process [76,77].

In field geophysical explorations, particularly within the vadose zone, Archie’s Law provides an empirical relationship between the bulk resistivity of a porous, water-saturated rock and its porosity and water saturation:

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

(6)

where:

• $\rho$: Bulk resistivity of the rock;
• a: Tortuosity factor (dimensionless);
• ${\rho }_w$: Resistivity of the pore water;
• $\varphi$: Porosity of the rock;
• m: Cementation exponent (empirical constant);
• $S_w$: Water saturation (fraction of pore space filled with water);
• n: Saturation exponent (empirical constant).

This relationship implies that as moisture content increases, resistivity decreases, making ERI effective for tracking moisture infiltration in soils. Variations in resistivity [78] thus reflect areas of high moisture content, which may correspond to weaker soil zones prone to failure [79].

3.2. ERI Testing in Soil and Rock

Though ERI tests are crucial for understanding the properties of both subsurface soil and rock, however, they differ fundamentally in their ERI application and interpretation due to the distinct differences in the physical characteristics of these two materials. The primary difference lies in the factors influencing the resistivity of soil and rock, and the methodologies used to interpret the resistivity measurements in soil versus rock contexts. In general, ERI testing is more applicable for geotechnical investigation in soil than rock, while in rock, it is often used for its structural integrity and hydrogeological assessments.

The resistivity of soil is significantly influenced by factors such as water content, porosity, and the degree of saturation. These factors are crucial because they affect the soil’s ability to conduct electricity, with higher moisture content generally leading to lower resistivity values [80,81]. Additionally, soil composition, including the presence of clay, can alter resistivity due to its conductive properties [82]. Resistivity in soils is highly sensitive to moisture variations, leading to significant temporal changes, whereas rocks, particularly igneous and metamorphic types, show more stable resistivity values unless fractured or weathered. In rocks, resistivity is more closely related to the intrinsic properties of the rock matrix, such as mineral composition and porosity. The presence of fractures and the type of pore fluids also play a significant role. Unlike soil, the resistivity in rocks is less affected by moisture content and more by the rock’s inherent properties [83]. Rocks generally exhibit higher resistivity values than soils due to lower moisture content and greater mineral compactness, while soils, especially clay-rich ones, tend to have lower resistivity due to higher water retention and ion mobility.

Electrical resistivity tests in soils often involve laboratory-based soil-box tests to determine resistivity under controlled conditions, which can be influenced by factors like density and temperature [84]. Field-based methods are also used for rapid subsurface profiling, but interpreting these results can be complex due to the heterogeneous nature of soils [85]. For rocks, resistivity tests are often integrated with other geophysical methods to map subsurface structures and properties. These tests are used to infer properties like strength and moisture content, which are crucial for geotechnical analysis [83]. During field investigations, soil typically provides better electrode contact, ensuring stable measurements, whereas rock surfaces may require conductive gels or drilled electrodes to improve contact and data reliability. Also, the depth of investigation differs in soil and rock. In soil, ERI testing usually focuses on shallow depths, while in subsurface rock formations, it is often used for deeper investigations for bedrock profiling, fault detection, or groundwater exploration. Accordingly, in soils, closely spaced electrode configurations are used that provide high-resolution data for near-surface characterization, while in rocks, wider electrode spacing is typically used to achieve greater penetration depth at the cost of resolution.

3.3. ERI-Based Approaches

ERI methodologies address various objectives, including moisture content mapping, identifying saturated zones and water tables, detecting fractures and slip surfaces, and time-lapse monitoring for dynamic changes. Each application targets specific aspects of subsurface instability, contributing to a comprehensive understanding of slope conditions.

3.3.1. Moisture Content Mapping

Moisture content variations significantly influence slope stability, as increased water content reduces soil matric suction and shear strength. ERI provides a robust framework for profiling moisture distribution by mapping resistivity variations, where low resistivity values typically indicate higher moisture content. This capability is particularly valuable for monitoring slopes after rainfall events or in regions with expansive soils.

Nobahar et al. [1] highlighted the use of ERI to detect variations in soil moisture, which is critical for understanding water infiltration patterns, drainage systems, and zones prone to saturation. These moisture variations can destabilize slopes by reducing soil cohesion and increasing the potential for failure.

Salunke et al. [34] used ERI to map moisture content variations in expansive clay highway embankments in Mississippi. Their study identified zones with high moisture content that correlated with subsurface water accumulation, which increased slope failure risks. This application provided valuable insights for proactive drainage design and mitigation strategies.

Linrong et al. [86] further demonstrated the use of ERI for profiling moisture content variations along the depth of embankments. This capability enables a layered understanding of moisture dynamics, improving predictions of slope stability under various environmental conditions.

Nobahar et al. [1] applied ERI to map subsurface moisture variations within embankments. Their work showed that low resistivity values were indicative of higher moisture content, while high resistivity suggested drier conditions. This profiling approach provided a clear understanding of how moisture changes influenced embankment stability, supporting targeted mitigation efforts.

By identifying zones of excessive moisture accumulation, ERI supports the proactive design of drainage and mitigation systems.

3.3.2. Detection of Saturated Zones and Water Tables

The presence of saturated zones and perched water tables is a primary contributor to slope failures, as saturated soils exhibit reduced shear strength. Electrical Resistivity Imaging (ERI) excels in identifying these zones by mapping resistivity contrasts between saturated and unsaturated conditions, providing critical insights into slope hydrological dynamics and potential failure mechanisms [87].

Nobahar et al. [1,88] demonstrated the effectiveness of ERI in detecting zones of full saturation and perched water tables, which are critical precursors to slope instability. Saturated zones, characterized by very low resistivity values, correlate strongly with increased moisture content, leading to reduced soil strength and a higher likelihood of failure along slip surfaces.

Solla et al. [35] applied ERI to identify perched water zones beneath highway embankments, showing how low resistivity values indicated areas of significant water saturation. Their findings underscored the heightened failure risks associated with these zones, particularly under load-bearing conditions.

Watlet et al. [89] used ERI to map the depth and extent of perched water tables formed after seasonal rainfall. These saturated zones, often acting as key triggers for slope movement, were found to significantly influence the stability of slopes during extreme weather events.

High-resistivity zones in ERI data often indicate unsaturated conditions, typically located near the surface or in well-drained areas. Linrong et al. [86] highlighted the importance of tracking resistivity changes over time to observe transitions from unsaturated to saturated conditions, signaling an increased risk of failure. This temporal monitoring capability allows for early intervention and mitigation measures.

Nobahar et al. [1] applied ERI to detect subsurface saturated zones in embankments, emphasizing how areas with very low resistivity values were directly linked to zones of high moisture that posed significant risks for embankment failure. Their study demonstrated the applicability of ERI in monitoring changes in moisture levels over time to assess slope vulnerability.

Ijari et al. [90] integrated ERI with in situ soil testing to measure subsurface properties, including Soil Matric Suction (SMS) and moisture variations at depths of 1.5 m, 3 m, and 4.5 m. ERI identified zones of high moisture that contributed to reduced shear strength and the development of potential slip surfaces. Additionally, their use of the Finite Element Method (FEM) simulated slope behavior under varying rainfall and groundwater conditions, providing comprehensive insights into failure mechanisms and aiding in defining the Factor of Safety (FS) for slopes.

By integrating ERI data with complementary tools such as field sensors and numerical models, the depth, extent, and location of perched water tables can be better understood. These insights support more accurate assessments of slope vulnerability under diverse environmental conditions, enabling proactive design and mitigation strategies.

3.3.3. Subsurface Fracture and Slip Surface Identification

Fractures and slip surfaces are critical precursors to slope failures, making their early detection essential for effective risk mitigation. The ERI is a powerful tool for visualizing subsurface resistivity anomalies that are often indicative of fractured zones or potential slip planes [91]. These insights are invaluable for assessing slope stability and identifying high-risk areas for proactive intervention.

Salunke et al. [34] demonstrated the integration of ERI with field sensor data to detect subsurface fractures and potential slip surfaces within expansive clay layers. Their study revealed that resistivity variations corresponded to zones of mechanical weakness, providing crucial insights into areas prone to future slope failures.

Similarly, Nobahar et al. [1] utilized ERI to identify slip surfaces in embankments, emphasizing the role of low-resistivity anomalies in indicating saturated or fractured zones that contribute to instability. The ability to delineate these features allowed for targeted slope reinforcement measures and improved failure prediction accuracy.

Watlet et al. [89] applied time-lapse ERI to monitor evolving subsurface fractures in slopes. By capturing changes in resistivity over time, their approach enabled the detection of progressive slip surface formation, particularly following heavy rainfall events. This dynamic monitoring capability enhances the ability to anticipate slope movement.

The integration of ERI with other geophysical techniques further enhances its utility. Solla et al. [35] combined ERI with Ground Penetrating Radar (GPR) to improve the resolution and accuracy of subsurface imaging. The complementary strengths of these methods enabled a more comprehensive understanding of subsurface conditions, including fracture networks and slip plane geometries.

Additionally, Ijari et al. [90] incorporated ERI data into Finite Element Method (FEM) simulations to model the mechanical behavior of slopes under various environmental conditions. By integrating ERI-derived fracture and slip surface data, their study provided actionable insights into slope failure mechanisms and informed the design of stabilization strategies.

By leveraging ERI in combination with other tools, such as GPR and numerical modeling, the detection and analysis of subsurface fractures and slip surfaces become more precise. These advancements contribute significantly to the development of early warning systems and the implementation of effective mitigation measures.

3.3.4. Detection of the Vadose Zone

The vadose zone, or unsaturated zone, is the soil layer above the water table, characterized by partially filled pores containing both air and water [92]. Its properties directly influence water infiltration, soil moisture dynamics, and slope stability—critical factors in assessing highway embankment health [86]. Understanding the vadose zone is essential for proactive slope monitoring, as moisture fluctuations in this zone can significantly affect slope stability, particularly in clayey soils prone to swelling and shrinkage.

The ERI is a key tool for investigating soil moisture content and heterogeneity within the vadose zone [21]. Variations in resistivity provide valuable insights into moisture patterns and saturation levels, enabling the identification of potential instability zones due to moisture changes. This capability is crucial for designing effective mitigation and intervention strategies [74].

Linrong et al. [86] demonstrated the application of ERI in characterizing the extent of the vadose zone and its influence on slope stability during wet and dry periods. Their findings emphasized how resistivity variations over time could provide a detailed understanding of moisture dynamics and the conditions contributing to slope instability.

Nobahar et al. [1] utilized ERI to characterize the vadose zone above the water table by tracking resistivity changes over time. Their work highlighted how moisture fluctuations within this zone significantly influenced the stability of slopes. The ability to monitor these changes allowed for the early detection of destabilizing conditions and informed proactive maintenance decisions.

For highway embankment monitoring, advanced methodologies such as sequential geostatistical approaches further refine moisture profiling within the vadose zone. Yeh et al. [92] introduced a stepwise inversion technique that incorporates stochastic models to account for spatial variability in moisture–resistivity relationships. This method updates resistivity measurements over time, creating an evolving moisture map that adapts to recent changes. Such dynamic mapping is critical for real-time monitoring of highway slopes, allowing for timely interventions and enhanced early warning systems.

The integration of ERI-derived moisture profiles with numerical and geostatistical models enhances the accuracy of predictions regarding moisture dynamics and their impact on slope stability. These insights are vital for proactive slope monitoring and the development of tailored mitigation strategies for different soil types and environmental conditions.

3.3.5. Time-Lapse Resistivity Monitoring for Dynamic Changes

Monitoring resistivity changes over time provides critical insights into the dynamic processes affecting slope stability [93]. Time-lapse ERI captures how subsurface resistivity evolves in response to environmental factors, such as rainfall, groundwater fluctuations [94], or drying periods [95]. This technique enables the early detection of failure risks by identifying trends of increasing saturation or resistivity changes that signal the drying or weakening of soils.

Alam’s work [96] demonstrates the effectiveness of time-lapse resistivity monitoring in understanding the hydrological behaviors of soil over time. A weekly examination of three evapotranspiration covers constructed on a slope was conducted, producing heatmaps that visually represented the subsurface resistivity dynamics over several weeks. The electrodes were initially spaced at 5 feet, enabling a broad mapping depth of 10 m. For enhanced resolution, a specific section of the slope (indicated by the arrows) was examined using electrodes spaced at 0.15 m, offering a mapped depth of 1 m, as shown in Figure 2a. This multiscale approach provided both a general overview and localized details of moisture dynamics.

On 6 June 2016, areas of low resistivity (depicted in cooler colors on the heatmaps) indicated high moisture content (Figure 2b). By 14 June 2016, these regions (outlined by the dashed rectangle) had vanished, suggesting that the moisture had either evaporated or drained away, as evidenced by increased resistivity values shown in Figure 2b. This transition highlights the soil’s drying process, which reduces the risk of shear strength loss due to excessive moisture.

Conversely, the consistent presence of deep blue patches in the heatmaps, representing stable areas of high moisture content, emphasizes zones of concern. If these patterns persist over time, they signal potential areas for slope instability. Under such circumstances, proactive measures such as excavation of the critical area and mitigation would be recommended before failure mechanisms could develop.

While weekly examinations provide granular insights into rapid moisture dynamics, prolonged examinations conducted over monthly intervals are often employed to balance time and cost implications. Such monthly monitoring remains effective for observing hydrological events and assessing long-term patterns of moisture accumulation or dissipation. These insights are invaluable for proactive slope failure mitigation and prevention, particularly in identifying areas that may require intervention before instability occurs.

Alam’s study [96] highlights the practical value of time-lapse resistivity monitoring in both short-term and long-term hydrological assessments. By tracking temporal and spatial variations in resistivity, practitioners can make data-driven decisions to design and implement effective slope stabilization measures. When integrated with predictive models and complementary monitoring tools, this methodology significantly enhances the reliability of early warning systems for slope failure.

3.3.6. Statistical Modeling and Machine Learning for Hydrological Predictions

The integration of statistical and Machine Learning (ML) methods [97,98] has proven to be effective for analyzing and predicting hydrological behaviors within the vadose zone, as demonstrated by Alam [96] in Figure 2. By leveraging field data, including electrical resistivity, Volumetric Moisture Content (VMC), and matric suction measurements, these methods enable enhanced understanding and forecasting of subsurface dynamics.

Statistical Modeling

The study conducted by Alam et al. [99] explored relationships between electrical resistivity and soil hydrological parameters such as VMC and matric suction extracted using multiple sensors. Resistivity versus VMC exhibited a nonlinear inverse relationship, where increasing VMC corresponded to reduced resistivity. This relationship was modeled using a sigmoidal function derived from the Van Genuchten soil–water characteristic curve (SWCC), as shown:

$$R_{\theta }(\theta ,X)=R_r+{\displaystyle \frac{R_s-R_r}{{[1+{(a/\theta )}^b]}^c}}$$

(7)

where:

• $R_{\theta }$: Electrical resistivity corresponding to VMC ($\theta$) and a parameter vector $X(R_s,R_r,a,b,c)$;
• $R_r$: Residual resistivity at dry conditions;
• $R_s$: Soil resistivity at saturated condition;
• $a,b,c$: Curve fitting parameters.

Similarly, the resistivity versus matric suction relationship was modeled using:

$$R_{\psi }(\psi ,Y)=R_r+{\displaystyle \frac{R_s-R_r}{{[1+{\left(p\psi \right)}^q]}^r}}$$

(8)

where:

• $R_{\psi }$: Resistivity related to matric suction ($\psi$) and a parametric vector $Y(R_s,R_r,p,q,r)$;
• $R_r,R_s$: Residual and saturated resistivities, respectively;
• $p,q,r$: Curve fitting parameters defining the shape of the function.

These models were validated using field data and demonstrated reliable prediction capabilities, offering insights into moisture infiltration and suction variations across seasons.

Field Insights

The graphical relationships derived in the study include the following:

• Resistivity vs. VMC: Figure 3 [99] demonstrates the decrease in resistivity as VMC increases, with a noticeable clustering of data points at low resistivity or high VMC values.
• Resistivity vs. Matric Suction: Figure 4 [99] indicates increasing resistivity with greater suction, reflecting the efficiency of soil resistivity in field-scale matric suction characterization.

These plots illustrate the critical thresholds for saturated and residual moisture, and illustrate the correlation of hydrological parameters with resistivity for developing models for predicting and evaluating slope soil behavior under varying hydrological conditions for proactive slope monitoring and intervention strategies.

Machine Learning Approaches

In addition to statistical models, ML techniques such as regression analysis and neural networks have been applied to predict subsurface behaviors dynamically. In [99], Alam et al. conducted a field investigation that highlighted the potential of machine learning frameworks to forecast resistivity changes based on meteorological inputs (e.g., precipitation and temperature) and sensor-based hydrologic data. Features derived from resistivity profiles, moisture sensors, and tensiometers could be fed into different ML models to predict temporal changes in moisture distribution, zones of potential saturation or drying, and subsurface instability indicators.

The combined use of statistical and machine learning models can provide an adaptive framework for analyzing time-lapse resistivity data to predict hydrological events and soil behaviors enabling early warnings for slope instability, improved drainage and mitigation planning, and cost-effective alternatives to traditional sensor-based monitoring systems.

3.4. Integration with Other Methods

The effectiveness of Electrical Resistivity Imaging is significantly enhanced when integrated with complementary geophysical and analytical methods. These integrations improve the depth of analysis, resolution, and real-time monitoring capabilities, offering a comprehensive understanding of slope stability and failure mechanisms. The following methods exemplify the synergy achieved through such integrations:

• GPR and ERI: GPR complements ERI by providing high-resolution imaging of shallow subsurface features, including fractures, moisture content, and stratification. GPR excels in detecting shallow slip surfaces and mapping soil composition. For instance, Solla et al. [35] combined GPR with ERI to improve subsurface image resolution, highlighting the strengths of using both techniques for identifying critical slip surfaces and structural anomalies and overcoming the limitations of individual geophysical methods.
• SRT and ERI: SRT provides additional insights into subsurface layering and soil strength properties, making it a valuable companion to ERI. By measuring P-wave and S-wave velocities, SRT characterizes the mechanical properties of subsurface materials, such as elastic moduli and shear strength. When combined with resistivity profiles obtained through ERI, these measurements offer a more comprehensive assessment of slope stability. Hussain et al. [5] and Sari et al. [100] emphasized the utility of integrating SRT and ERI for landslide early warning systems, particularly in near-real-time applications.
• LiDAR and ERI: LiDAR technology provides precise surface measurements, which, when integrated with ERI’s subsurface insights, create a holistic view of structural conditions. This combination enables the capture of both surface deformations and subsurface anomalies, improving the analysis of slope stability. Sreenath et al. [44] demonstrated the potential of combining LiDAR and ERI to analyze structural conditions comprehensively, particularly in capturing subtle surface deformations and their relationship to underground instability.
• FEM and ERI: Numerical modeling using the FEM is frequently combined with ERI data to simulate slope behavior under various environmental conditions, such as rainfall and seismic events. ERI-derived resistivity data parameterize FEM models, incorporating soil properties like moisture content, shear strength, and elastic modulus. Solla et al. [35] utilized FEM simulations to assess failure risks under different moisture and seismic conditions. Moreover, Nobahar et al. [1] leveraged Plaxis 2D FEM software to perform back-calculations, identifying slip surfaces and evaluating soil strength parameters based on UAV and ERI data.

3.5. Limitations of ERI Application

Electrical resistivity Imaging (ERI) is a valuable geophysical method for subsurface exploration, but it has several limitations, including spatial resolution, electrode spacing, interpretation challenges in heterogeneous terrains, and sensitivity to environmental conditions. These limitations can affect the accuracy and reliability of ERI data, necessitating careful consideration and methodological adaptations to mitigate their impact on subsurface characterization.

Spatial Resolution: The spatial resolution of ERI is influenced by a variety of factors, which include geological conditions, electrode configurations, data processing techniques, and noise levels. These factors collectively determine the quality and accuracy of the subsurface images produced by ERI. Understanding these influences is crucial for optimizing ERI applications in geotechnical infrastructures’ health monitoring. Below are the primary factors affecting ERI spatial resolution:

• Geological Conditions

  -

  Resistivity Contrast: The resolution is better in low-resistance strata compared to high-resistance ones. If a high-resistance layer is beneath a low-resistance layer, the resolution capability declines [101].

  -

  Layer Depth and Thickness: Variations in layer depth and interlayer thickness can affect resolution, with deeper layers generally having lower resolution [101]. The depth of investigation also affects resolution, with deeper layers being more challenging to resolve accurately.

• Electrode Configuration

  -

  Array Type: Different electrode arrays, such as dipole–dipole, bipole–bipole, pole–bipole, pole–pole, and gradient arrays, offer varying levels of sensitivity and resolution. The choice of array impacts spatial resolution significantly [102]. Therefore, selecting the appropriate array type is crucial.

  -

  Electrode Spacing: Shorter electrode spacing, especially with invasive electrodes, can enhance the resolution but may introduce errors if not properly managed [103]. Smaller electrode spacing results in a denser set of apparent resistivity measurements, enhancing the resolution of shallow subsurface features. This is particularly important for identifying compacted soil areas, which often occur near the surface. However, this approach is more time-consuming and logistically challenging [104]. Larger electrode spacing allows for deeper penetration into the subsurface, which can be beneficial for identifying features at greater depths. However, this comes at the expense of reduced resolution for shallow layers, which can be problematic when the primary interest is in near-surface compaction [104,105]. Increasing electrode spacing can lead to a deterioration in the accuracy of the inverted resistivity images, affecting both resistivity distribution and the delineation of interfaces between different soil layers. This degradation is more pronounced when the electrode spacing exceeds the thickness of the top subsurface layer [104].

• Data Processing and Inversion Techniques

  -

  Inversion Algorithms: Traditional algorithms may smooth out details, reducing resolution. Integrating deep learning with prior physical information can enhance resolution by providing a more detailed inversion model [106].

  -

  Regularization and Sensitivity Analysis: Proper regularization and sensitivity analysis are essential for improving resolution by minimizing inversion artifacts [107].

• Noise and Measurement Strategies

  -

  Noise Levels: Increased noise can lead to random jumps in resistivity profiles, degrading resolution. Effective noise management is crucial for maintaining high-quality images [101].

  -

  Measurement Strategies: Transitioning from 2D to 3D Imaging and optimizing measurement strategies can improve the amount of information obtained, thus enhancing spatial resolution [108].

Interpretation in Heterogeneous Terrain: The presence of heterogeneous terrain significantly impacts the interpretation of ERI data by introducing complexities that can lead to erroneous results. This is primarily due to the variations in surface characteristics and subsurface structures that affect resistivity measurements. Proper planning and execution of ERT surveys, along with advanced modeling techniques, are essential to mitigate these challenges and enhance the accuracy of geological interpretations.

• Impact of Terrain Heterogeneity on ERI Data

  -

  Surface Complexity: Heterogeneous terrains, such as those with limestone outcrops or coarse blocky surfaces, can cause significant variations in resistivity data. These variations can lead to ambiguous interpretations if not properly accounted for during the design, field implementation, and data processing stages [109,110].

  -

  Electrode Contact Issues: In rugged terrains, achieving optimal galvanic contact between electrodes and the ground is challenging, which can affect data quality. Innovative solutions like conductive textile electrodes have been developed to improve contact efficiency and reduce survey time in such environments [110]. However, they induce additional costs of investigation.

  -

  Topographical Effects: Irregular topography in the earth infrastructure system can distort ERI data, necessitating terrain corrections or inversion techniques that incorporate topography to obtain reliable subsurface images. These methods help isolate the true subsurface response from topographical influences [111,112].

Sensitivity to Environmental Conditions: Environmental circumstances significantly impact the accuracy of electrical resistivity measurements, influencing the interpretation of subsurface variabilities. Variations in humidity, temperature, and other environmental factors can introduce noise and errors, affecting the reliability of resistivity data. Understanding these influences is crucial for improving measurement accuracy and model interpretation.

• Environmental Sensitivity

  -

  Humidity: Humidity changes can alter the resistivity of subsurface soil, as demonstrated in studies where resistivity exhibited a logarithmic relationship with absolute humidity. This is due to water molecule adsorption on mineral surfaces within material pores, affecting the entire sample rather than just the surface [113].

  -

  Temperature Variations: Temperature fluctuations can cause significant errors in resistivity measurements. For instance, in environments with fluctuating temperatures, such as those exposed to solar radiation, errors can be reduced by using improved sensors that compensate for temperature differences [114]. Temperature compensation is essential for accurate subsurface characterization, as resistivity can be over- or underestimated without accounting for temperature changes [115].

  -

  Soil Properties: The physical properties of soils, such as moisture content, negative pore water pressure, and density, affect resistivity. Variations in these properties lead to different resistivity values in loose versus dense conditions, as seen in studies of clayey silt and silty sand [116]. In hyper-arid environments, the lack of moisture presents challenges for resistivity measurements, requiring robust modeling to account for low conductivity conditions [117].

  -

  Contact Resistance and Noise: Contact resistance and measurement noise are higher in dry conditions, leading to increased data discarding and model errors. This noise can cause discrepancies between observed and predicted resistivity data, affecting model accuracy [118].

While ERI faces these limitations, advancements in technology, advanced numerical modeling, data processing techniques, and methodology are available nowadays to address some of these challenges. For example, the use of deep learning algorithms and unconventional electrode arrays can enhance data interpretation and resolution capabilities [119,120]. Heterogeneous terrain poses challenges to ERT data interpretation; however, it also offers opportunities for innovation in investigation techniques and data processing. The development of new electrode technologies and integrated geophysical approaches can significantly enhance the reliability of ERI surveys in complex terrains. Environmental variabilities can introduce significant challenges to resistivity measurements and data accuracy. However, advancements in sensor technology and modeling techniques offer solutions to mitigate these effects. Improved sensors and compensation methods can enhance measurement accuracy, allowing for more reliable interpretations of subsurface structures. Despite all these limitations, ERI remains one of the powerful subsurface characterization techniques for monitoring and assessment of geotechnical infrastructure health.

3.6. Summary of ERI Approaches

The key methodologies and applications of Electrical Resistivity Imaging (ERI) discussed in this section are summarized in

Table 2. Summary of ERI approaches for subsurface monitoring.

Approach	Capabilities	Applications	Limitations
Time-Lapse Monitoring	Tracks temporal changes in subsurface resistivity to track moisture variations and seepage.	Monitoring seasonal variations, detecting early signs of slope instability, and assessing drainage efficiency.	Limited by cost and time for frequent data acquisition; requires advanced processing for high temporal resolution.
Vadose Zone Analysis	Maps soil moisture dynamics in the unsaturated zone above the water table.	Predicting infiltration pathways, monitoring swelling and shrinking of expansive soils, and early detection of moisture-driven slope failures.	Highly dependent on soil type and resolution of electrode spacing.
Saturated Zone Identification	Detects areas of high moisture or perched water tables through low resistivity signatures.	Identifying zones of reduced shear strength, assessing failure risks in slopes, and designing drainage solutions.	Limited depth penetration in conductive soils like clays; requires complementary geophysical methods for deeper characterization.
Fracture and Slip Surface Detection	Visualizes resistivity anomalies associated with fractures or slip planes in the subsurface.	Locating potential failure planes, improving landslide prediction, and validating numerical models.	Limited resolution for small-scale features; enhanced by integrating with GPR or seismic methods.
Statistical Modeling	Quantifies relationships among resistivity, volumetric moisture, and matric suction.	Predicting soil hydrological behavior; developing early-warning thresholds.	Requires robust field data for validation; sensitive to parameter calibration.
Machine Learning (ML)	Automates analysis and predicts resistivity changes based on environmental, climatic, and hydrological data.	Time-series analysis, real-time failure prediction, and moisture dynamics modeling.	Accuracy depends on data volume and quality; computationally intensive for large datasets.
GPR and ERI Integration	High-resolution imaging of shallow features complements resistivity profiles.	Detecting shallow slip surfaces, fractures, and stratification.	Limited penetration depth; enhanced when paired with resistivity for deep insights.
SRT and ERI Integration	Combines resistivity data with seismic velocities for mechanical property analysis.	Slope stability assessment, identifying slip planes, and soil strength evaluation.	Requires precise calibration and interpretation; sensitive to environmental noise.
LiDAR and ERI Integration	Merges surface topography with subsurface resistivity insights.	Comprehensive slope monitoring, capturing both surface deformation and subsurface anomalies.	Affected by vegetation or atmospheric conditions; higher processing needs.
InSAR and ERI Integration	Provides high-resolution ground deformation data alongside subsurface insights.	Monitoring cumulative deformations over time; identifying subsidence patterns.	Vegetation and phase decorrelation can affect accuracy; benefits from combining with UAV-based InSAR.
Integration with Numerical Models (e.g., FEM)	Provides resistivity data to parameterize subsurface properties for slope stability simulations.	Modeling slope behavior under various environmental scenarios and designing mitigation strategies.	Requires accurate calibration of input data and integration with field observations.

.

Table 2 highlights the diverse capabilities and applications of ERI, showcasing its importance in understanding subsurface behavior and contributing to comprehensive slope stability assessments.

3.7. GPR and SAR for Subsurface Monitoring

Ground Penetrating Radar (GPR) and Synthetic Aperture Radar (SAR) are essential remote sensing techniques that complement Electrical Resistivity Imaging (ERI) in subsurface investigations. GPR operates by transmitting high-frequency electromagnetic waves into the ground and analyzing the reflected signals to detect subsurface anomalies, such as fractures, voids, and moisture variations. It is particularly effective for identifying shallow geological features and buried structures, making it valuable for infrastructure assessment and geotechnical investigations. Synthetic Aperture Radar (SAR), on the other hand, utilizes microwave signals to map surface and subsurface deformations with high spatial resolution. UAV-mounted SAR systems enable continuous monitoring of ground displacement over time, detecting subtle changes indicative of subsidence, landslides, and structural instability. By integrating GPR and SAR data with ERI, a comprehensive multi-sensor approach can be achieved, enhancing subsurface characterization and improving predictive modeling for slope stability and infrastructure monitoring [121].

4. Data Fusion Framework

Effective monitoring of transportation infrastructure requires integrating surface and subsurface data to provide a comprehensive view of structural health [122]. UAV-based surface data and ERI-derived subsurface insights offer complementary perspectives, with UAVs capturing high-resolution imagery of surface conditions and ERI mapping moisture and resistivity profiles in the subsurface. Data fusion frameworks bridge these datasets, enabling holistic analysis critical for infrastructure health monitoring [123].

4.1. Data Fusion Levels

Data fusion methodologies operate at different stages of data processing, as illustrated in Figure 5. The data fusion process begins with integrating raw datasets, such as UAV and ERI data, which are transformed into informative datasets through data-level fusion. This step preserves essential details while preparing the data for further analysis [124].

Second, the feature-level fusion process extracts and refines representative features, such as displacement, moisture content, and resistivity, enabling the combination of high-quality features from diverse sources. Finally, decision-level fusion aggregates predictions from multiple models, resulting in robust decision-making frameworks critical for infrastructure health assessments.

Adapting this framework to UAV and ERI datasets highlights the multidimensional nature of data fusion. As shown in Figure 5, the three fusion levels are described as follows:

• Data-Level Fusion: Raw datasets from UAVs, ERI, and other sensors are combined into a unified dataset, preserving essential details while simplifying processing.
• Feature-Level Fusion: Specific features, such as UAV-derived surface deformation patterns and ERI-detected subsurface moisture anomalies, are extracted and integrated using dimensionality reduction techniques like Principal Component Analysis (PCA) and clustering algorithms [123]. Wang et al. [124] highlighted how this improves slope stability predictions by merging high-quality features from diverse datasets.
• Decision-Level Fusion: Predictions from multiple models are aggregated to create robust frameworks for decision-making. Ensemble techniques, such as Random Forests or stacking algorithms [125], enhance classification accuracy and support predictive maintenance.

One of the core challenges in Structural Health Monitoring (SHM) is inspection frequency, as infrastructure conditions change dynamically over time. Traditional inspection methods are costly, labor-intensive, and infrequent, limiting the ability to track deterioration effectively.

The integration of UAVs and ERI within a data fusion framework addresses key challenges in geotechnical infrastructure inspection, including accessibility, data accuracy, and the need for frequent updates. UAVs, combined with ERI, enhance inspection frequency, data collection efficiency, and accuracy, enabling continuous, high-resolution monitoring that surpasses conventional methods. UAV automation facilitates frequent, cost-effective, and minimally invasive inspections, ensuring early detection of structural deterioration and improving infrastructure assessments. By fusing UAV-acquired surface data with ERI subsurface imaging, this approach provides comprehensive, high-resolution insights that significantly enhance damage detection, monitoring efficiency, and maintenance prioritization. This integration supports informed decision-making, resource optimization, and proactive interventions, ultimately improving the resilience and longevity of geotechnical infrastructure. Below, we discuss some major challenges in geotechnical infrastructures inspection.

• Accessibility and Safety: Traditional inspection methods often require physical access to hazardous or remote areas, posing safety risks to inspectors. UAVs can safely access these areas, reducing the need for human presence in dangerous locations [126,127].
• Data Collection Frequency: Frequent inspections are necessary to ensure infrastructure safety, but traditional methods are time-consuming and costly. UAVs can automate data collection, allowing for more frequent and efficient inspections [128,129].
• Data Accuracy and Consistency: Manual inspections can be subjective and inconsistent. UAVs equipped with advanced sensors provide high-resolution, objective data, improving the accuracy and reliability of inspections [130,131].

4.2. Practical Experimental Data Collection and Integration

Data fusion frameworks combine multi-dimensional datasets from UAVs, ERI, and additional sensing technologies to create a unified understanding of structural health [132]. For instance, integrating UAV optical imagery with ERI resistivity data allows precise correlation of surface deformations with subsurface conditions, enhancing failure prediction capabilities. Multi-sensor approaches leverage UAV’s high spatial resolution and ERI’s depth profiling to generate comprehensive risk assessments [133].

Advanced frameworks incorporate complementary methods such as GPR and SAR to extend monitoring capabilities [121]. GPR offers shallow subsurface detection, while SAR monitors large-scale surface deformations over time, creating a synergy that enhances both spatial resolution and temporal dynamics. By combining these methods, data fusion frameworks provide actionable insights into slope stability and structural integrity.

4.3. Applications of Machine Learning in Data Fusion

Machine Learning (ML) algorithms play a pivotal role in enhancing data fusion frameworks by automating multi-source data analysis and enabling predictive modeling. For instance, Salunke et al. [34] employed XGBoost (XGB) [134] and Support Vector Regression (SVR) models to predict soil moisture content from UAV imagery and ERI datasets, with XGB demonstrating superior predictive accuracy. Similarly, Nobahar et al. [1] utilized Random Forest (RF) and Support Vector Machines (SVMs) to classify embankment stability as stable, marginally stable, or unstable based on features such as resistivity values, surface temperature variations, and historical rainfall data.

Time-series models, such as Long Short-Term Memory (LSTM) networks [135], are increasingly being used to forecast hydrological patterns. Watlet et al. [89] applied Hierarchical Agglomerative Clustering (HAC) to time-series resistivity data for identifying zones with distinct hydrological behavior, while LSTM-based models have successfully captured temporal dependencies, predicting moisture infiltration and its implications for slope stability.

Incorporating hybrid deep learning models, such as Convolutional Neural Networks (CNNs) combined with LSTMs, further enhances the detection of complex structural patterns. Dang et al. [136] demonstrated such integration in a data fusion approach for Structural Health Monitoring (SHM), where diverse signal processing techniques, such as Discrete Wavelet Transform (DWT) and Empirical Mode Decomposition (EMD), were integrated into a 3D tensor for robust damage detection.

Beyond predictive modeling, automation techniques in image analysis significantly enhance UAV-based infrastructure monitoring. The integration of Convolutional Neural Networks (CNNs) with generative adversarial networks enhances the robustness of pavement distress detection, achieving an average precision of 78.2% [137]. The CNNs and deep learning models are widely used for crack detection on pavements and bridges [138], feature extraction from LiDAR point clouds, and vegetation stress assessment in multispectral imagery. Unsupervised clustering methods, such as K-means and DBSCAN, facilitate automatic segmentation of deformation patterns, while semantic segmentation models improve defect identification in high-resolution UAV images [139]. Additionally, automated interferogram processing in InSAR enables real-time subsidence monitoring, reducing manual effort and increasing detection accuracy [140]. These automation techniques streamline data interpretation, improve efficiency, and enable scalable, real-time monitoring solutions.

4.4. Real-Time Monitoring and Anomaly Detection

Real-time data integration is crucial for early-warning systems. IoT-enabled sensors [141,142] combined with UAV and ERI data streams facilitate continuous monitoring, enabling real-time anomaly detection and proactive decision-making. Salunke et al. [34] utilized ML-based anomaly detection systems with UAV thermal imagery to provide automated alerts for slope instability. These systems are particularly valuable for high-risk infrastructure like embankments adjacent to highways and dams, where timely interventions can prevent catastrophic failures [32].

Predictive algorithms analyze live data streams to identify early signs of instability [143]. Integrating these algorithms with real-time data from IoT-enabled sensors allows practitioners to implement automated alerts and mitigation strategies, thereby enhancing infrastructure resilience [144,145].

4.5. Applications in Slope Stability Monitoring

Data fusion frameworks have proven instrumental in slope stability monitoring [146,147]. Researchers can map deformation patterns, track moisture infiltration, and predict potential slip surfaces by combining UAV-derived surface imagery with ERI subsurface profiles. Hao et al. [28] extended this approach by integrating UAV and satellite imagery to estimate biomass and assess vegetation health, further enhancing slope stabilization strategies. Below, we discuss the most critical data fusion applications of UAV and ERI.

• Comprehensive Data Integration for Infrastructure Monitoring: UAVs with high-resolution cameras and LiDAR provide detailed surface imaging, while ERI detects subsurface anomalies such as water infiltration, fractures, and unstable soil layers [148]. This has been particularly useful in landslide-prone areas where early detection of slope instability is crucial. For example, UAV-based RGB imaging has been used in highway embankment monitoring, while ERI has identified moisture accumulation zones that contribute to failures [61,149].
• Enhanced Monitoring of Dams, Bridges, and Roads: UAV-ERI data fusion has been successfully applied in monitoring dams and bridge foundations where surface-level degradation and subsurface seepage pose structural risks. A case study on rubble mound breakwaters demonstrated that UAV-based thermal imaging detected surface cracks, while ERI revealed deeper water infiltration zones, preventing failure [150]. Similarly, UAV-based LiDAR and ERI fusion have been used in airport runway assessments, identifying subsurface voids that could lead to pavement collapse.
• Predictive Maintenance and Risk Mitigation: In urban areas, UAV-ERI fusion has been integrated with AI-driven risk assessment tools to predict infrastructure deterioration patterns [127,151]. Machine learning algorithms analyze UAV-derived surface deformations and ERI-based moisture profiles to predict potential failure points in highway slopes and underground tunnels [124,152]. This enables early interventions, reducing long-term maintenance costs.
• Improving Decision-Making for Disaster Response: UAV and ERI data fusion has been instrumental in post-earthquake and flood damage assessments, allowing engineers to quickly identify unstable infrastructure before reopening roads and bridges [153]. UAV-based aerial damage assessment combined with ERI-driven subsurface imaging allows targeted repairs, reducing downtime and optimizing resource allocation [154,155,156].

Additionally, integrating ML-driven models with Finite Element Method (FEM) simulations enables dynamic analysis of slope behavior under environmental stressors like rainfall or seismic activity. Nobahar et al. [1] utilized FEM simulations parameterized with ERI data to predict failure mechanisms and identify critical slip planes. Such models allow for proactive maintenance by simulating potential failure scenarios and optimizing mitigation strategies.

4.6. Summary of Data Fusion Approaches

Table 3. Summary of data fusion approaches for infrastructure monitoring.

Approach/Method	Capabilities	Applications	Limitations
Data-Level Fusion/Inspection	Combines raw data from multiple sources like UAVs, ERI, GPR, and SAR to create unified datasets.	Comprehensive aggregation for slope stability monitoring and risk assessment.	Requires high-quality datasets; prone to dimensionality issues with large-scale integration.
Feature-Level Fusion/Inspection	Extracts and integrates features (e.g., surface deformations, subsurface moisture anomalies) from UAV and ERI datasets using techniques like PCA and clustering.	Enhances prediction accuracy for slope stability; supports classification of weak zones.	Sensitive to feature selection; requires advanced algorithms to avoid loss of critical information.
Decision-Level Fusion/Inspection	Aggregates outputs from multiple models (e.g., Random Forests, stacking) to enhance decision-making.	Used for early-warning systems and classification of risk levels (e.g., stable, marginally stable, unstable).	Dependent on the reliability of individual models; challenges in harmonizing conflicting outputs.
GPR and ERI Integration/Inspection	Combines high-resolution shallow subsurface imaging with resistivity profiles for enhanced analysis.	Detecting fractures, slip surfaces, and shallow water infiltration zones.	Limited penetration depth; requires complementary methods for deeper insights.
LiDAR and ERI Integration/Inspection	Combines precise surface measurements from LiDAR with subsurface resistivity data.	Holistic slope monitoring by capturing surface deformations and subsurface anomalies.	Affected by atmospheric and environmental factors; computationally intensive.
InSAR and ERI Integration/Inspection	Merges precise deformation monitoring from InSAR with resistivity-based subsurface analysis.	Long-term deformation analysis, subsidence detection, and monitoring of slope stability.	Challenges include phase decorrelation and vegetation interference; limited spatial resolution.
Machine Learning and ERI Integration/Analysis	Uses ML algorithms (e.g., Random Forest, XGB, LSTM) to predict hydrological behaviors and subsurface instability.	Predicting slope movement, moisture infiltration, and structural health.	Requires large, high-quality datasets; sensitive to overfitting and computational demands.
Hybrid Deep Learning/Analysis	Combines time-series data with deep learning (e.g., CNN-LSTM) for robust anomaly detection and forecasting.	Advanced slope monitoring, moisture prediction, and failure risk identification.	Computationally demanding; dependent on noise-free data for accurate predictions.
IoT and ERI Integration/Inspection	Combines IoT-enabled sensors with UAV and ERI data for real-time monitoring.	Proactive slope stability management, automated alerts for slope instability.	Integration challenges; requires reliable network infrastructure for real-time data streams.
GIS and ERI Integration/Inspection	Integrates geospatial tools (e.g., ArcGIS) with ERI profiles to generate 3D models.	Mapping failure zones, visualizing deformation trends, and risk classification.	Limited by geospatial data resolution and processing power for large datasets.
FEM and ERI Integration/Analysis	Parameterizes FEM simulations with resistivity data to model slope behavior under stressors.	Simulating failure mechanisms, optimizing intervention strategies.	Computationally expensive; dependent on accurate resistivity and soil properties.

summarizes the data fusion approaches discussed in this paper, categorized by fusion levels (data, feature, decision) and the integration of imaging and sensing techniques such as GPR, LiDAR, and InSAR with ERI. The table highlights the practical applications of these approaches, such as slope stability monitoring, real-time anomaly detection, and predictive risk assessment.

Data-level fusion aggregates datasets from UAVs, ERI, and IoT sensors. Feature-level fusion refines critical parameters like surface deformation and subsurface moisture, while decision-level fusion enhances predictive accuracy using ensemble models. Advanced technologies, including FEM, GIS, and hybrid machine learning models, further extend these frameworks’ capabilities for infrastructure health monitoring.

5. Open Research Gaps and Future Directions

This article explores the integration of various technologies and methodologies for slope stability monitoring, including UAV imaging, Electrical Resistivity Imaging (ERI), LiDAR, and advanced data fusion frameworks. While significant advancements have been achieved, several challenges and research gaps remain, such as deeper subsurface data acquisition, automation in data fusion, and urban-specific monitoring frameworks.

Table 4. Summary of reviewed papers and their utilized imaging approach. SZ: Saturated Zones and water tables; FS: Subsurface Fracture and Slip Surface Identification; MP: Moisture Profiling; VZ: Vadoze Zone; FI: Field Instrumentation; FEM: Finite Element Method; GIS: Geographic Information System; ML: Machine Learning.

Ref.	Infrastructure	Surface Research Approach	Subsurface Research Approach	Data Fusion Technique
[1]	Slope	UAV (RGB, Thermal)	ERI (SZ, FS)	FEM (Analysis)
[25]	Slope	UAV (RGB)	ERI (SZ, FS)	None
[34]	Pavement	UAV (Infrared, Optical)	None	ML (SVR, XGB) (Analysis)
[89]	Slope	UAV (RGB, Thermal)	ERI (SZ)	None
[1]	Slope	UAV (LiDAR, RGB)	ERI (SZ, FS)	GIS (Inspection)
[5]	Slope	UAV (Thermal, Multispectral)	ERI	None
[32]	Slope	UAV (InSAR)	None	None
[1]	Slope	UAV (RGB)	ERI (SZ, FS)	FEM (Analysis)
[31]	Pavement	UAV (RGB, Infrared)	None	None
[157]	Slope	UAV (RGB)	ERI (SZ, FS)	None
[88]	Slope	FI, UAV (RGB, Thermal)	ERI (MP, SZ)	None
[99]	Slope	None	ERI (MP, SZ)	ML (Analysis)
[158]	Slope	FI	ERI (MP, SZ, VZ)	None
[100]	Dam	UAV (RGB, LiDAR)	GPR	FEM (Analysis)
[86]	Slope	UAV (RGB, Multispectral)	None	None
[44]	Pavement	UAV (RGB, Infrared)	ERI (SZ)	None
[35]	Slope, Pavement, Dam	UAV (RGB, Thermal, InSAR)	GPR	FEM (Analysis)
[24]	Slope	UAV (RGB)	ERI (SZ)	None
[159]	Slope	UAV (Thermal)	ERI (SZ)	None
[92]	Slope	UAV (RGB, Infrared)	ERI (MP)	ML (RF) (Analysis)
[160]	Infrastructure	UAV (InSAR, RGB, Multispectral)	ERI (SZ)	GIS (Inspection)
[136]	Slope	UAV (Infrared)	ERI (FS)	ML (Hybrid Deep Learning) (Analysis)
[28]	Infrastructure	UAV (RGB, LiDAR)	None	ML (Biomass Modeling) (Analysis)
[90]	Pavement	UAV (RGB, Multispectral)	None	ML (XGB, PCA) (Analysis)
[124]	Infrastructure	UAV (RGB)	None	Data-Level Fusion (Inspection)
[36]	Bridge	UAV (RGB, Infrared)	ERI	None

presents a synthesis of some of the reviewed papers, highlighting the infrastructure types, surface and subsurface approaches implemented, and the data fusion techniques, showing the diverse methodologies employed and identifying trends and gaps in the field.

The analysis in Table 4 reveals the predominance of UAV-based imaging techniques, such as RGB and LiDAR, for surface monitoring, complemented by ERI for subsurface profiling. However, integrating advanced data fusion methodologies, particularly ML models, remains underexplored in urban environments.

We identified various gaps in the literature and future research directions, including the following:

• Enhancing the depth of subsurface monitoring through integrated geophysical techniques such as ERI and Seismic Refraction Tomography (SRT).
• Automating data fusion frameworks using advanced machine learning algorithms to streamline analysis and improve prediction accuracy.
• Developing tailored urban-specific slope stability models that incorporate surface, subsurface, and anthropogenic factors.
• Expanding real-time monitoring capabilities through IoT-enabled sensors and multi-source data integration.

By addressing these gaps, researchers can significantly advance the field of slope stability monitoring, enabling proactive risk mitigation and designing more resilient infrastructure systems. The remainder of this section discusses these research gaps and future directions separately.

5.1. Slope Monitoring in Urban Areas

Despite advancements in slope monitoring technologies, several research gaps persist, particularly in urban contexts. These gaps highlight the challenges of addressing complex urban infrastructure dynamics and underscore the need for innovative solutions.

Urban slopes face unique stressors, including concentrated runoff, impermeable surfaces, and construction activities, which are often overlooked in generalized models. Developing urban-specific slope models incorporating subsurface, surface, and anthropogenic data is vital for accurate risk assessment and slope management [90].

5.1.1. Deep Subsurface Data

Accurately characterizing deeper subsurface layers remains a critical challenge in urban slope monitoring. ERI, while effective for shallow investigations, struggles with depth penetration in highly conductive soils, such as clays. Existing studies have shown the potential of integrating more profound geophysical methods like GPR and SRT to overcome these limitations [1]. In regions with complex lithologies, such as the Karakoram Himalaya, combining ERI with GPR has demonstrated improved mapping of bedrock interfaces and subsurface fractures, essential for slope stability assessments [86].

5.1.2. Long-Term Monitoring

Temporal changes in slope stability due to factors like rainfall, freeze–thaw cycles, and groundwater fluctuations have not been studied in urban settings. Automated systems like PRIME and time-lapse ERI (4D ERT) provide promising frameworks for monitoring these changes over extended periods [89]. However, their adoption in urban environments remains limited due to challenges adapting to diverse soil types and anthropogenic influences. Incorporating rainfall, soil moisture sensors, and long-term monitoring setups is essential to address this gap effectively.

5.1.3. Automated Data Fusion Frameworks

Existing monitoring systems often rely on manual data integration from UAV, ERI, and other geophysical techniques, which is inefficient and prone to errors. The application of ML, including CNN and LSTM models, can automate data fusion processes and enhance predictive analytics. For example, ML models have been shown to improve classification accuracy for slope stability based on UAV-derived imagery, ERI profiles, and environmental data [34,136]. Expanding ML-driven automation for real-time monitoring in urban embankments is a critical research priority.

5.1.4. Vegetation Interference and Surface Occlusions

Dense vegetation and urban surface features pose challenges to UAV-based photogrammetry. Advanced imaging techniques, such as LiDAR and multispectral imaging, can address these limitations by providing high-resolution DTMs and penetrating vegetation cover [44]. Such tools are critical for accurately mapping urban slopes with dense vegetation or complex structural overlays.

5.1.5. Real-Time Early Warning Systems

Real-time early warning systems remain underdeveloped in urban contexts. While IoT-enabled sensors and predictive algorithms can provide near-instantaneous alerts, integrating diverse datasets—such as UAV imagery and ERI data—into real-time systems presents challenges in data standardization and processing efficiency [34]. Developing integrated and real-time solutions is crucial for proactive slope management.

5.1.6. Multidisciplinary Integration

The integration of geophysical methods, such as ERI and UAV data, with numerical modeling techniques like FEM is underutilized. Coupling these methods with geospatial tools, such as GIS, allows for holistic assessments of slope behavior under environmental and seismic conditions [5]. Enhanced multidisciplinary approaches can provide actionable insights for urban slope stability.

5.2. Proposed Future Research Directions

To address these gaps, the following research directions are proposed:

• Enhanced Data Fusion Techniques: Develop data fusion frameworks that integrate UAV, InSAR, ERI, and LiDAR datasets using ML models like CNNs and ensemble ML methods to produce comprehensive 3D slope stability maps.
• Real-Time Monitoring Systems: Expand real-time monitoring capabilities by combining IoT-enabled sensors with UAV and geophysical imaging systems, supported by predictive analytics using LSTM networks.
• Urban-Specific Applications: Incorporate urban-specific datasets, such as drainage networks and underground utilities, into slope monitoring frameworks using GIS tools.
• Multi-Sensor Integration: Build systems that combine UAV photogrammetry, InSAR, ERI, and LiDAR data with advanced geospatial analysis tools for improved monitoring accuracy and predictive modeling.
• Dynamic Seismic Interactions: Develop models that incorporate the effects of seismic activity alongside long-term environmental factors for real-time urban slope stability assessments.
• Predictive Models for Moisture-Suction Dynamics: Advance the relationship between ERI-derived resistivity and matric suction for better predictive modeling in expansive urban soils.
• Data Standardization: Establish standardized protocols for data acquisition and processing to improve interoperability across various monitoring techniques.