Towards Operational Dam Monitoring with PS-InSAR and Electronic Corner Reflectors ^†

Abstract

Dams are crucial for ensuring water and electricity supply, while also providing significant flood protection. Regular monitoring of dam deformations is of vital socio-economic and ecological significance. In Germany, dams must be constructed and operated according to generally accepted rules of engineering. The safety concept for dams based on these rules relies on structural safety, professional operation and maintenance, safety monitoring, and precautionary measures. Rather time-consuming in situ techniques have been employed for these measurements, which permit monitoring deformations with either high spatial or temporal resolution, but not both. As a means of measuring large-scale deformations in the millimeter range, the Differential Interferometric Synthetic Aperture Radar (DInSAR) technique of Persistent Scatterer Interferometry (PSI) is already being applied in various fields. However, when considering the operational monitoring of dams using PSI, specific characteristics need to be considered. For example, the geographical location of the dam in space, as well as its shape, size, and land cover. All these factors can affect the visibility of the structure for the use with PSI and, in certain cases, limit the applicability of SAR data. The visibility of dams for PSI monitoring is often limited, particularly in cases where observation is typically not feasible due to factors such as geographical and structural characteristics. While corner reflectors can improve visibility, their large size often makes them unsuitable for dam infrastructure and may raise concerns with heritage protection for listed dams. Addressing these challenges, electronic corner reflectors (ECRs) offer an effective alternative due to their small and compact size. In this study, we analyzed the strategic placement of ECRs on dam structures. We developed a new CR Index, which identifies areas where PSI alone is insufficient due to unfavorable geometric or land use conditions. This index categorizes visibility potential into three classes, presented in a ‘traffic light’ map, and is instrumental in selecting optimal installation sites. We furthermore investigated the signal stability of ECRs over an extended observation period, considering the Amplitude Dispersion Index (ADI). It showed values between 0.1 and 0.4 for many dam structures, which is comparable to normal corner reflectors (CRs), confirming the reliability of these signals for PSI analysis. This work underscores the feasibility of using ECRs to enhance monitoring capabilities at dam infrastructure.

1. Introduction

Dams, broadly categorized into embankment dams and gravity dams, are critical infrastructure supporting water management and fulfilling key roles in water supply, flood control, power generation, and tourism. They ensure consistent river flows, enable recreational activities, and provide essential water storage capacities, particularly in response to the growing challenges posed by climate change. The socio-economic and ecological importance of dams, however, is juxtaposed with the catastrophic risks associated with their failure, making reliable and continuous monitoring imperative [1,2]. Germany exemplifies the rigorous approach to dam safety through its stringent standards, particularly DIN 19700-10:2004-07 [3]. These standards mandate tailored monitoring programs that address the specific needs of each dam. Traditional techniques, such as tacheometry and pendulum measurements, have long been employed for tracking vertical and horizontal deformations with high precision [4].

However, these methods are labor-intensive and typically conducted only at one position on the dam or only twice a year, limiting their capacity to detect transient or sudden deformations that may indicate potential structural issues. Advancements in satellite-based remote sensing technologies, specifically DInSAR, and as an advanced form of DInSAR, the Persistent Scatterer Interferometry (PSInSAR) technique, have revolutionized dam monitoring [5]. PSI leverages radar signals from multiple satellite passes to detect deformations with millimeter-level accuracy. Its capacity to provide dense spatial coverage and frequent temporal updates makes PSI an invaluable tool for infrastructure monitoring. Freely available radar data, such as Copernicus Sentinel-1 imagery, have enabled the widespread application of PSI, as demonstrated by services like the German Ground Motion Service (BBD) and the European Ground Motion Service (EGMS) [6,7,8].

Despite its transformative potential, PSI’s effectiveness varies based on structural and environmental factors, leading to inconsistencies in the density and quality of persistent scatterer (PS) points. This variation is particularly evident between embankment and gravity dams. Embankment dams, often covered with vegetation, may exhibit reduced radar reflectivity, whereas gravity dams, though generally more reflective, can face challenges related to orientation, size, or surrounding terrain. To address these challenges, the CR Index was developed as a predictive tool to evaluate PSI feasibility for various land cover classes at a large regional scale [9,10,11]. By integrating geometric parameters (e.g., slope and radar incidence angle) with land use data, the CR Index enables a tailored assessment of PSI applicability. The CR Index has proven effective in predicting PS density distributions across varying terrains, making it a valuable resource for site-specific monitoring strategies [11]. This research adapts the CR Index specifically for dam monitoring on the local scale of complex infrastructure, allowing a more precise evaluation of both embankment and gravity dams. This approach ensures that even less observable structures can be assessed effectively.

Following this analysis, the installation of electronic corner reflectors (ECRs) is explored as a complementary strategy to enhance PSI observability. As the only commercially available option, ECRs from MetaSensing are smaller, lighter, and easier to install than traditional passive reflectors, making them ideal for complex structures like dams, as passive corner reflectors are often too big for dams, or the issue of monument protection does not allow installation. ECRs actively emit strong intensity signals in C-Band, ensuring better visibility in radar imagery, even in challenging environments. Additionally, ECRs can be remotely operated and are less intrusive, making them suitable for use on protected or historically significant structures. They are easy to maintain, and power can be supplied via cable or solar panel. Depending on the weather and outside temperature, they operate for up to 2/3 weeks on battery power alone. ECRs act as artificial PS points, providing consistent radar signal reflections, even in challenging environments with insufficient natural scatterers. Guided by CR Index evaluations, ECR deployments offer a targeted solution to extend PSI-based monitoring to previously under-observed structures. The combination of PSI, the CR Index, and ECRs represents a significant step forward in dam monitoring. This integrated framework addresses the limitations of natural PS variability and establishes an operational observation strategy that is robust, scalable, and adaptable to various dam types and environments. The following sections will detail the methodologies, findings, and implications of this study, highlighting the potential of these techniques to redefine dam monitoring standards.

2. Materials and Methods

2.1. Background and State of the Art

Advancements in DInSAR have enabled the identification and characterization of deformations that affect dams. Specifically, the Persistent Scatterer Interferometry technique, or PSI [12,13], provides deformation estimations with millimeter accuracy. PSI has been successfully used to monitor various embankment dams [14,15,16,17,18,19].

TerraSAR-X data were used to monitor the Plover Cove Dam in Hong Kong between 2008 and 2012, while other studies utilizing Cosmo-SkyMed and Envisat analyzed dams in Italy, Uzbekistan (Charvak Dam), and China (Three Gorges Dam) [20,21,22]. Sentinel-1 data facilitated the monitoring of deformation at the La Viñuela Dam in Spain and the Mosul Dam in Iraq, where PSI provided a crucial alternative due to the infeasibility of on-site measurements caused by water intrusion into evaporite layers [15,18,23,24]. PSI has also been applied to assess ground movement around hydraulic infrastructure such as dikes and levees, as seen in studies of embankments along the Guadalhorce River, enhancing flood risk prediction [23].

The feasibility of PSI for embankment and gravity dams was analyzed in [25], serving as a guideline for future research. Other hydraulic structures, such as locks, have also been successfully monitored using PSI, exemplified by the Hessigheim lock in Germany, where X-band data revealed a high density of PS points and an annual deformation cycle [26]. Further PSI-based dam monitoring studies are discussed in [27,28]. Additionally, advanced radar interferometry techniques, including SBAS-InSAR and PS-InSAR, have been increasingly utilized for monitoring dam stability, covering in-construction dams [29], reservoir dams [30,31,32], and tailings dams [33,34,35]. Comparisons between PSI and SBAS methodologies further highlight their effectiveness for displacement monitoring and risk management in various dam structures [36].

The selection of case studies is vast; however, most of the studies mentioned refer to embankment dams. There are a few exceptions that also precisely investigate gravity dams [4]. A closer examination of the dam structures discussed in the literature reveals that not every structure is equally observable. Challenges such as dense vegetation, steep topography, and small structural footprints can reduce the density and quality of persistent scatterers. To compensate for these conditions and achieve higher observability from traditionally less observable areas, corner reflectors (CRs) are increasingly being utilized [37,38]. These devices provide fixed, highly reflective PS points that enhance radar backscatter in areas with low natural reflectivity.

Electronic corner reflectors (ECRs), in some cases also referred to as compact active transponders (CATs), have increasingly been used to enhance the reliability and accuracy of Synthetic Aperture Radar (SAR) applications. These devices serve as artificial scatterers, in areas with limited natural reflectivity. Some studies have demonstrated the utility of ECRs [39,40,41], however, not in relation to dams.

The MetaSensing ECR-C Transponder is a compact and efficient device designed to enhance radar observability for Persistent Scatterer Interferometry (PSI) applications. With a size of 360 × 570 × 233 mm and a weight of 10.2 kg, it features a PVC radome for durability. The transponder operates with low power consumption, requiring less than 5 W in active RF mode and a maximum of 15 W when charging via AC or 65 W when using solar panels. It is powered by a LiFePO4 main battery (12.8 V, 10 Ah), with an additional LiPo support battery (3.7 V, 0.6 Ah). Power can be supplied via AC (100–240 VAC) or solar panels (65 Wp, max input 27 V), making it adaptable for remote installations.

For communication and data transmission, the device supports USB (Type-A 2.0) and Wi-Fi (802.11 b/g/n), ensuring flexible connectivity for remote monitoring and control. The RF characteristics of the ECR-C include a broad bandwidth exceeding 100 MHz in the C-band (around 5.4 GHz), making it suitable for satellites such as Sentinel and RADARSAT. It provides an internal RF gain of 50 dB, with a P1dB compression point at 20 dBm, ensuring high signal stability. The antenna gain is 15 dBi, with an azimuth beamwidth of 20° and an elevation beamwidth of 40°, allowing for effective signal alignment across different orbital paths. The transponder supports both linear horizontal and vertical polarization, which can be selected through its GUI software. The Effective Isotropic Radiated Power (EIRP) reaches a maximum of 35 dBm, although typical operational levels are around 5 dBm.

ECRs are particularly advantageous in optimizing signal return and improving the identification of persistent scatterers (PS) in challenging observational environments. This study further explores their scalability and adaptability for various infrastructure monitoring scenarios, underscoring their versatility for both small-scale and large-scale applications [39]. This study specifically provides in-depth insights into the operational mechanisms of ECRs, focusing on their ability to overcome limitations associated with conventional corner reflectors. It highlights their active nature, which allows for optimized reflectivity, even under suboptimal geometric conditions, and discusses their utility in settings where traditional methods face constraints, such as in densely vegetated areas or complex terrain. This makes ECRs a transformative technology for SAR interferometry, with promising implications for infrastructure and environmental monitoring [39]. Additionally, early experiences combining ECRs and Sentinel-1 data reveal significant improvements in radar backscatter stability and precision, validating ECRs as an essential tool for long-term deformation monitoring [39].

The integration of ECRs with PSI and other SAR techniques has demonstrated significant potential to overcome the limitations posed by low natural reflectivity or unfavorable geometric configurations. These studies underscore the role of ECRs as indispensable tools for advancing the precision and reliability of SAR-based monitoring systems. This highlights one of the novel aspects of this study: the installation of ECRs on dams to enhance visibility in radar images and facilitate a planned subsequent PSI analysis.

2.2. Methods

Before discussing the processing in detail, the following sections will first outline the study area, and the data used. Based on these points, the modification of the existing CR Index and the criteria for installing the ECRs will finally be explained. The general workflow for this study, from the investigation of PSI visibility to the handling, installation, and maintenance of the ECRs is a systematic process designed to ensure their optimal performance and long-term reliability. As shown in the workflow in Figure 1, the process begins with site selection and preparation. These steps are critical to ensure that the ECRs are placed in locations that maximize radar visibility while minimizing environmental interference, such as vegetation or topographic shadowing. To achieve this, a modified CR Index [10] is applied, which combines indices that evaluate geometric suitability (R Index) and land cover characteristics (Land Use Index). This approach allows for the identification of optimal installation sites, ensuring that the reflectors provide consistent and reliable signals for subsequent PSI analyses, while also accounting for practical considerations such as accessibility and operational feasibility.

Once the sites are prepared, the next phase involves the deployment of the ECRs, which includes their installation and alignment. This process ensures that the reflectors are properly positioned to optimize their functionality and align with the radar’s incidence angles for consistent backscatter signals. Key considerations during deployment include structural stability, environmental characteristics, and the establishment of reliable power and data transmission systems. These steps are critical for ensuring the long-term operability of the ECRs and their ability to support subsequent PSI analyses. A detailed explanation of each phase is provided in the later sections.

2.2.1. Study Area and Data

For this research project 8 dams of different types were considered, which all are situated in the jurisdiction of the Ruhrverband in North Rhine-Westphalia (Germany). The Ruhrverband is an organization based in Germany specifically tasked with managing water resources in the Ruhr River Basin [42]. In Figure 2, a general overview of the Ruhrverband’s area of jurisdiction is provided, showing the eight major dam structures and their location in the western German state of North Rhine-Westphalia.

All data used for the processing of the CR Index as well as for later validation purposes are listed in

Table 1. Datasets used in this study, their source, and the main use in this study.

Data	Data Source	Use
Sentinel-1 SLC	[43]	Processing R Index (single scene), backscatter analysis
LiDAR elevation data (1 m)	[44]	DEM generation, processing R Index
ATKIS landcover data	[46]	Processing CR Index
OSM landcover data	[47]	Processing CR Index
BBD data	[7,48]	Validation CR Index
Sentinel-2 imagery	[49]	Validation CR Index

. Sentinel-1 C-Band data are valuable for detecting surface movements with radar interferometry techniques, enabling cost-effective monitoring of large areas [43]. Additionally, LiDAR data from the Geodata Infrastructure NRW (GDI NRW) with 1 m resolution offer high-precision terrain models that are instrumental in creating detailed topographic maps [44]. This 1 m digital elevation model (DEM) was used for further CR Index processing and preprocessing of Sentinel-1 data.

Figure 2. (a) Jurisdiction of the Ruhrverband [42] in relation to the metropolitan area of Duisburg, Essen, Dortmund, and Soest. All eight major dam structures of the jurisdiction are depicted: Möhne, Sorpe, Henne, Ennepe, Verse, Fürwigge, Lister, and Bigge. Locations for later ECR deployment are marked with red arrows. (b) The jurisdiction’s location in Germany [45].

Figure 2. (a) Jurisdiction of the Ruhrverband [42] in relation to the metropolitan area of Duisburg, Essen, Dortmund, and Soest. All eight major dam structures of the jurisdiction are depicted: Möhne, Sorpe, Henne, Ennepe, Verse, Fürwigge, Lister, and Bigge. Locations for later ECR deployment are marked with red arrows. (b) The jurisdiction’s location in Germany [45].

The ATKIS land cover data, also from GDI NRW, provide detailed information on land cover across North Rhine-Westphalia, supporting environmental planning and change assessment around dam areas [46]. Their use is further explained in the following sections. They were combined with OpenStreetMap (OSM) data, which contribute extensive geographic details, enhancing the geographical context of the study by integrating local infrastructure [47].

The Ground Movement Service of Germany (BodenBewegungsdienst Deutschland—BBD) supplies detailed measurements of ground displacement, which are crucial for monitoring geological changes and assessing the stability of areas surrounding dam structures over time [7,48]. These data were used to validate the CR Index. For this purpose, Sentinel-2 data were also used, offering additional optical imagery for validation, ensuring the accuracy and reliability of the analysis [49].

The characteristics of the Sentinel-1 data used are listed in

Table 2. Attributes of the Sentinel-1 data in this study [43].

Data	Sentinel-1 Ascending	Sentinel-1 Descending
Number of scenes	20	21
Temporal resolution	12	12
Acquisition mode	Interferometric Wideswath (IW)	Interferometric Wideswath (IW)
Polarization	VV	VV
Wavelength	C-Band	C-Band
Relative orbit number	15	139
Frame	164	421

. The SLC (Single Look Complex) data are particularly important as they are necessary for the generation of the CR Index, as well as for subsequent validation and backscatter analysis of the installed ECRs. In the case of the backscatter analysis, a time series of 20 scenes for the ascending direction and 21 scenes for the descending direction is used. However, for processing the CR Index, a single scene is sufficient, as only metadata is required.

2.2.2. CR Index Calculation and Modification

The CR Index is implemented as described in [10]. It is an index that provides information about the probability of identifying PS points in a specific study area. The CR Index incorporates not only geometric information derived from a DEM and Sentinel-1 geometry as the R Index does, but also information about land use and land cover. If land use information has not been included in the previous step, the term R Index is used.

The CR Index is generally processed following the methodology outlined in [10]. The key adaptation and resulting novelty lie in the development of a Land Use Index (LUI) specifically tailored for Germany. While [11,50] provide examples for Great Britain and Italy, their values cannot be directly applied due to the differing nomenclature of land cover classes in the ATKIS system, although they serve as initial benchmarks. Consequently, values derived from the preliminary study [51] using BBD data were employed for the calculation of the LUI. These values are based on BBD PS point densities for the individual land use classes defined in ATKIS. This approach is also necessary because ATKIS offers a significantly more detailed subdivision into individual subclasses compared to similar classifications found in the literature, such as those in [10,11,51]. The calculation of the CR Index is followed by validation using data of the German Ground Motion Service.

The R Index estimates potential PS density based on topography and radar recording configuration, with values ranging from −100 to 100 [10]. High R Index values indicate optimal conditions for PSI technology, influenced by elevation, slope orientation and inclination, and the Sentinel-1 incidence angle. The index is calculated using parameters such as slope inclination, local slope orientation (adjusted to satellite flight paths), and radar incidence angles derived from Sentinel-1 data.

Figure 3 illustrates the general workflow for CR Index computing. Slope and aspect data are extracted from a 1 m resolution DEM after converting it to GeoTIFF format and reprojecting it. Slope inclination directly contributes to the R Index, while slope orientation requires aspect correction for alignment with satellite geometry.

Sentinel-1 data preprocessing involves orbit correction, conversion to Ground Range Detected (GRD) format, and extraction of incidence angle information, which represents the angle between the radar signal and the ellipsoid surface (WGS 84). This is accomplished using SNAP software during geocoding, with the DEM serving as a critical input for accurate R Index calculation.

R Index calculation begins with the correction of slope orientation (aspect) maps based on Sentinel-1 satellite heading angles relative to the north axis. These heading angles, derived from Sentinel-1 metadata, are averaged across scenes captured in three swaths. It is important to note that polarization (VV or VH) does not affect this adjustment.

The next step involves preparing the necessary input parameters, which include slope inclination, corrected slope orientation (aspect), and Sentinel-1 incidence angles. These parameters are derived from the DEM and Sentinel-1 metadata. Slope and aspect are processed using tools like QGIS, where the aspect is corrected for radar geometry and all values are converted from degrees to radians to ensure compatibility with the calculation process.

The R Index is then calculated after [10], incorporating slope inclination, the corrected slope orientation, and radar incidence angles for each radar flight direction (ascending or descending). The resulting R Index maps quantify the geometric conditions necessary for successful PS observation, highlighting areas where PSI techniques are most effective only based on geometrical factors.

To extend the R Index into the CR Index, land cover information is additionally incorporated. This is achieved here using ATKIS data from the digital basic landscape model provided by the Federal Agency for Cartography and Geodesy [46], supplemented by OpenStreetMap (OSM) data [47]. While OSM data are primarily used to identify power transmission towers (data layer: Power) [47], which are absent in the ATKIS dataset, they serve as excellent PS objects. This composite of the two datasets results in a significantly more detailed classification into various subclasses, as the land use component hereby features a significantly more detailed subdivision.

The CR Index is derived as a combination of the R Index and the Land Use Index (LUI) [10,11]. The LUI quantifies the likelihood of identifying PS points based on land use categories. Urban areas with a high density of artificial structures, such as buildings or power transmission towers, are assigned to higher LUI values, reflecting the improved PSI applicability in such environments. Conversely, areas dominated by vegetation receive lower LUI values. These processes are visualized in the lower section of the workflow in Figure 3. ATKIS data play a crucial role in assigning LUI values to land use classes. Two components are used for this: first, comparative values from the existing literature [10,11,51] as a first general approach; and second, custom LUI values derived from previous processing of the German state of Thuringia using GMS data [51].

These values, specific to ATKIS land use classes, are listed in

Table 3. Listing of all values for the Land Use Index (LUI) based on [11,51].

Class	LUI	Class	LUI
Rail traffic	50	Historic building or facility	100
Railway facility	75	Industrial and commercial area	65
Waterbody structure	100	Agriculture	0
Transportation structure	80	Moor	0
Industrial and commercial building or facility	30	Square	30
Sports, recreation, and leisure building facility	10	Maritime traffic (structures)	100
Mining operation	40	Other buildings or facilities	100
Rock, boulder, rock needle	0	Sports, recreation, and leisure area	10
Area of special functional character	100	Standing water (masked)	0
Mixed-use area	80	Road traffic	25
Stream (masked)	0	Swamp	0
Air traffic	20	Open pit, mine quarry	10
Aviation facility	10	Wasteland, vegetation-free area	10
Cemetery	10	Storage tank, storage structure	100
Copse	0	Forest	0
Harbor basin	0	Residential area	100
Spoil tip	0	OSM Power	100
Heath	0

and contribute to a more detailed and to a Germany-adapted CR Index calculation, as the original number of 7 land use classes [10] was expanded to 35 classes. The LUI is derived from PS point densities based on data accessible through the BBD web service. For this purpose, the BBD data were intersected with land use data to determine the number of available PS points for each land use class. Using the total area information for each class, a PS density was calculated and subsequently normalized into an LUI with a value range from 0 to 100 [51]. The resulting values are shown in Table 3.

For power transmission towers and urban areas, a maximum LUI of 100 was applied, since these areas show the highest PS densities. In these cases, LUI values from the literature [10,11] could be directly validated. The same LUI values were calculated, particularly for the urban, water, forest, and agricultural classes. However, for less common ATKIS classes such as moors or swamps, new values based on BBD data could be calculated, enhancing the concept of the existing LUI from [10,11]. Almost no PS are present in these areas, resulting in a LUI of 0 [51], similar to the finding of the reference studies [10,11]. These classes were rated lower than grassland or agriculture due to the influence of water. The results of all these findings from the calculation using BBD data are systematically shown in Table 3.

Similarly, the “buildings in water areas” class, which includes dam structures, required adjustments as it also encompasses locks, harbor facilities, and hydroelectric power stations. A critical step here is a visual inspection of the ATKIS classification. Misclassifications within ATKIS data, such as dams categorized as roads (e.g., Lister dam), present a key challenge and must be addressed in order to guarantee a correct analysis.

Roads generally reflect radar signals away from the sensor, reducing their suitability for PSI analysis. However, objects like streetlights or guardrails in road areas often generate PS points, warranting an LUI of 25 rather than 0, which was calculated in [51]. Conversely, dams classified as “structures in water areas” require further adjustments due to their vegetation cover, which reduces PSI applicability compared to concrete or stone surfaces.

A finer subclassification of ATKIS land use classes compared to other datasets necessitated additional fine-tuning. Misclassifications and generalizations must be carefully addressed, particularly for dam structures, where coverage varies significantly. Some are fully vegetated, while others are partially covered with concrete or stone, influencing their PSI potential. This highlights the critical importance of accurate LUI creation and classification adjustments for reliable PSI analysis. This additional fine-tuning step must be manually performed using expert knowledge, which significantly complicates the automation of processing, particularly in the case of dam structures.

Following the calculation of the newly adapted LUI, CR Index processing was performed according to [10,11]. It is a simple average of the R Index and the LUI, with the LUI incorporating a weighting factor. The final CR Index served as a critical factor in determining the potential need for ECRs for PSI analysis in badly observable areas and subsequently determining the optimal locations for the ECR installations, which is explained in the following section.

In a further step, a validation of the processed CR Index was utilized in similar manner to [51]. Since BBD data for Thuringia were already used for the LUI, another dataset from the state of North Rhine-Westphalia was used. By calculating BBD PS point densities across the study area, a direct assessment of the CR Index was achieved. The new CR Index adapted to Germany and its validation results enabled the division of the CR Index into three distinct classes, visualized on a traffic light map, for dam operators in Germany.

2.2.3. Site Selection and ECR Deployment

All eight dam structures of the Ruhrverband, including four gravity dams (Möhne, Fürwigge, Lister, Ennepe) and four embankment dams (Sorpe, Bigge, Verse, Henne), were evaluated as potential targets for ECR installation, based on the CR Index calculation. The primary selection criteria were the CR Index and traffic light maps, which assess radar observability. Structures with favorable or moderate CR Index values were prioritized, while those with poor conditions, such as the Fürwigge dam (due to radar shadowing), were excluded.

The Möhne dam was selected because it serves as a long-term study site and could benefit from precise ECR point definition. Similarly, Sorpe Reservoir was included to enhance point stability despite already having some PS points. In contrast, the Henne Reservoir, with abundant natural PS points, did not require additional observation via ECRs.

The BBD (Ground Motion Service Germany) data validated the traffic light maps and highlighted structures like Bigge, Verse, and Lister dams, which had sparse PS points, suggesting the need for improvement through ECR installation. These reservoirs exhibited only moderately high CR Index values, supporting their selection.

Practical considerations, including power supply, installation, maintenance, and protection against vandalism, were also evaluated. Limited space for solar panels posed challenges at dam walls, necessitating alternative power solutions. These factors, combined with the need to enhance PSI applicability, led to the selection of two dam walls (Möhne and Lister) and three dams (Sorpe, Bigge, Verse) for ECR observation.

For the Bigge dam, its large structure with distinct slopes required two ECRs—one on each side—to improve radar visibility and expand the PSI network.

Figure 4 highlights the key factors influencing ECR installation. ECRs are generally placed near the crest to avoid shadowing and ensure visibility from both ascending and descending radar directions. Placement at the center of the crest is preferred, as it corresponds to areas of maximum expected movement. However, obstacles like vegetation, fences, or terrain shadows must be considered, as they can impact visibility.

The proximity to power sources is critical, particularly for ECRs powered by cables. While vegetation can interfere with visibility, it also provides a natural barrier against vandalism. At Sorpe and Bigge Reservoirs, for example, ECRs are effectively concealed by hedges and fences. Additionally, LTE routers, which provide WiFi access via mobile networks, are located nearby to enable remote operation of the ECRs. VPN tunnel connections allow external control of the devices.

ECRs are installed horizontally, requiring customized mounting solutions. For embankment dams, a concrete foundation supports Plexiglas mounts for the ECRs. For dam walls, where space is limited and the vandalism risk is higher, cantilever arms with Plexiglas panels were used to secure the ECRs.

Once the complete setup of the ECRs is finalized and sufficient data is collected, additional analyses with the ECRs were conducted. This analysis focused on key statistical parameters, such as the standard deviation, mean and Amplitude Dispersion Index (ADI), to characterize the temporal stability of the radar signal and assess the ECR performance with regard to a PSI analysis [12]. Once initial PSI analyses can be conducted with the ECRs, the focus will also be on comparing ascending and descending orbits. Further questions relate to the quality and quantity of the new PS points obtained, which will be analyzed at a later stage of the project. Additionally, due to the dominance of the ECRs in the amplitude image, the possibility of existing natural PS points being obscured is worth considering, especially for structures like Möhne dam or Sorpe dam.

3. Results

This section presents the results in two parts: firstly, the CR Index values are reported and the traffic light maps are shown; and secondly, the results of the subsequent operation of the deployed ECRs are presented.

3.1. CR Index

Figure 5 illustrates the calculated CR Index for the selected dam walls, including Möhne and Lister, in both ascending and descending directions. Water surfaces were masked out during the processing to ensure accurate analysis of the dam structures. The CR Index values for dam walls are generally higher compared to embankment dams due to the absence of vegetation and their well-defined geometric characteristics. Möhne and Lister dams consistently exhibit high CR Index values, indicating their suitability for the PSI technique. The influence of acquisition geometry is clearly visible, with CR Index values differing significantly between the ascending and descending directions. Additionally, the differences between the airside and waterside of the dam walls are well-defined in the CR Index maps, with the waterside often displaying slightly lower values due to the interaction of radar signals with water and slope orientation.

Similar to the R Index, high CR Index values (ranging between 0 and 100) indicate a high probability of identifying PS points. Negative values, which occur less frequently in the CR Index than in the R Index, are associated with areas where the R Index exhibited strong negative values. These are often caused by unfavorable geometric conditions, such as radar shadowing or areas with obstructive vegetation.

The CR Index distribution for embankment dams, including Sorpe, Bigge, and Verse, displays similar variability depending on the imaging direction, as is shown in Figure 6. Overall, the CR Index values for the embankment dams are lower compared to those for the dam walls, primarily due to denser vegetation cover and the natural materials used in their construction. Despite this, the structural details of the dams are distinctly captured, with features such as slope steps and surface textures clearly visible in the CR Index maps. The dam crest stands out with consistently higher values, reflecting its favorable geometry and sparse vegetation, making it a critical area for monitoring. The satellite images included for each dam provide a visual reference for understanding the spatial context and structural layout. They facilitate direct comparison with the CR Index maps and help highlight key features such as crests, slopes, and adjacent landscapes.

The formation of a traffic light map in three classes in this study is intended to simplify the application of the CR Index for operators of dam facilities who have no experience in the use of radar remote sensing or similar methods. The visualization tool of the traffic light map offers an intuitive option for decision making and aims to help answer the following questions: Which dams or areas of specific dams are not observable at all? Which dams are observable even without ECRs? For which dams could an ECR provide an enhancement in observability. In order to provide an answer to this problem, class boundaries were defined on the basis of the validation of the CR Index, as shown in Figure 7.

The CR Index classification is divided into three primary classes, alongside a radar shadow category for negative values [10]. Areas with CR Index values under -20 are considered unobservable due to the absence of PS points, often caused by unfavorable geometric conditions (e.g., radar shadow). The lowest class (CR Index: 0 to 20) includes regions that are challenging to monitor using PSI. These areas often show low R Index or LUI values or both. While some PS points may still emerge—such as those resulting from individual streetlights or masts—these points arise from scatterers not adequately captured in existing ATKIS or OSM datasets.

Medium observability (CR Index: 20 to 50) is characteristic of areas with mixed conditions, such as zones with moderate scatterer coverage or suboptimal geometric configurations. In these areas, PSI observability is possible, though it may be enhanced significantly with the use of ECRs. Examples include areas with partial scatterer coverage, such as urban squares or mining zones, where PS points are still detectable despite moderate conditions.

High observability (CR Index: above 50) is observed in structured or urban areas with optimal geometric and land use conditions. These regions exhibit minimal discrepancies between ascending and descending orbits, as the LUI plays a dominant role in determining observability. This finding aligns with earlier studies [10,11,50], which also reported strong PSI suitability in areas with high CR Index values. Although high CR values exist for Bigge and Sorpe, they are spatially limited, making the use of ECRs beneficial for achieving better results. In close cooperation with engineers from the Ruhrverband, suitable locations for ECR installation were identified using the CR Index.

3.2. ECR Operation

Figure 8 shows an initial image of the ECR amplitudes in a radar image (GRD) for Bigge dam. Both ECRs can be recognized as typical corner reflector structures. Additionally, the northeastern ECR with a solar panel installed at Bigge dam is shown.

The bottom of the figure shows an example for Möhne dam. Here, too, the ECR is clearly visible in the amplitude image.

The ECRs were put into operation at the beginning of 2023, but initially, only the two Bigge ECRs in the ascending direction provided successful transmissions. In the descending direction, successful transmissions were only recorded from the third or fourth acquisition onward, and for Sorpe, not until the end of February, or the sixth acquisition. This is reflected in Figure 9 by significantly lower backscatter values (Sigma-0). In April, most ECRs also experienced weather-related failures. Apart from these incidents, the ECRs provide high backscatter values. A first temporal analysis (Figure 9) of radar backscatter signals reveals distinct movements for ascending and descending observations at the Bigge 1, Bigge 2, Lister, and Sorpe dams. These trends highlight the influence of observation geometry and environmental conditions on the performance of ECRs. The ascending mode demonstrates stable performance for the Bigge 1 and Bigge 2 reflectors, with Sigma-0 values consistently ranging between 5 and 6 dB. This reflects robust and reliable signal returns over time. Lister stands out with the highest average Sigma-0 values, peaking at approximately 11 dB, indicating optimal alignment and favorable site conditions in the ascending direction. In contrast, the Sorpe dam exhibits significant fluctuations, with Sigma-0 values occasionally dropping below 0 dB, particularly in early 2023 and late spring. Descending observations show greater variability across all sites compared to the ascending mode. Bigge 2 achieves high peak Sigma-0 values exceeding 13 dB, but the dynamic range, spanning from 3.74 to 13.97 dB, reflects sensitivity to changing conditions that are unknown so far. Sorpe displays consistently strong backscatter during favorable periods, with a mean Sigma-0 of 13.16 dB, although instability is evident during certain intervals. Lister performs poorly in descending mode, with occasional Sigma-0 values dropping below 0 dB. This underperformance suggests challenges specific to descending geometry at this site, which may be related to the water surface or the wall itself.

Results of the statistical analysis are shown in

Table 4. Results of the statistical analysis for both ascending and descending directions.

Direction	Dam	Count	Mean [dB]	Std [dB]	Min [dB]	Max [dB]	ADI
Ascending	Bigge 1	20	5.16	0.62	3.86	6.18	0.12
	Bigge 2	16	5.32	0.42	4.71	6.24	0.07
	Lister	14	10.95	0.74	9.7	12.15	0.06
	Sorpe	13	2.48	1.3	-0.1	4.62	0.52
Descending	Bigge 1	19	7.04	2.38	3.12	11.76	0.34
	Bigge 2	15	8.42	3.09	3.74	13.97	0.36
	Lister	15	3.11	2.87	-0.95	8.02	0.92
	Sorpe	14	13.16	1.67	8.53	15.24	0.12

. They include the mean, standard deviation, and range of Sigma-0 values, alongside the Amplitude Dispersion Index (ADI) to assess signal stability.

In the ascending mode, the performance of the reflectors varies across different sites. Bigge 1 and Bigge 2 exhibit stable and consistent backscatter signals, with mean Sigma-0 values of 5.16 dB and 5.32 dB, respectively. Both reflectors also have low standard deviations (0.62 for Bigge 1 and 0.42 for Bigge 2), further indicating reliable signal returns over time. Lister demonstrates the highest overall performance in ascending mode, with a mean Sigma-0 of 10.95 dB and a narrow standard deviation of 0.74. This reflects excellent stability and favorable conditions for signal reflection at this site.

In contrast, Sorpe performs the poorest among the reflectors in ascending mode, with the lowest mean Sigma-0 value of 2.48 dB and the highest variability, as indicated by a standard deviation of 1.30. These metrics highlight significant inconsistencies in Sorpe’s signal behavior. In the descending mode, the reflectors display more variability compared to the ascending observations. Bigge 2 emerges as a strong performer, with a mean Sigma-0 of 8.42 dB. However, this is accompanied by a high standard deviation of 3.09, indicating considerable variability in signal stability. Sorpe, while achieving the highest mean Sigma-0 of 13.16 dB in descending mode, also demonstrates a broad range of values, pointing to potential instability under certain conditions. Lister, on the other hand, performs the weakest in descending mode, with a mean Sigma-0 of only 3.11 dB and the highest standard deviation of 2.87. This highlights substantial challenges in maintaining consistent backscatter signals at Lister in descending observations, which are likely influenced by local site conditions or orientation factors.

The Amplitude Dispersion Index (ADI) values, as presented in the table, provide key insights into the stability and variability of radar returns from different dams under ascending and descending acquisition geometries. In the ascending direction, Lister and Bigge 2 exhibit the lowest ADI values (0.06 and 0.07, respectively), indicating stable and consistent signal returns. Sorpe, on the other hand, shows a notably higher ADI (0.52), reflecting significant variability in radar returns. In the descending direction, Sorpe demonstrates an impressive stability with a low ADI value of 0.12, while Lister shows a strikingly high ADI value of 0.92, highlighting considerable signal variability. Bigge 1 and Bigge 2 show moderate ADI values (0.34 and 0.36, respectively), reflecting relatively stable, yet slightly variable, radar returns.

4. Discussion

In this section, the results presented in the previous section are discussed with a particular focus on two central aspects: the validation of the CR Index and the operation of electronic corner reflectors, including the associated challenges. A preliminary small study had already addressed the identification of suitable locations for ECRs using the CR Index for two dams. However, this is further expanded here through extensive validations [52]. Furthermore, the development of the traffic light map and the analysis of backscatter signals are additional extensions to [52]. These aspects are analyzed in relation to the study’s objectives, highlighting their implications for the overall research findings and their relevance within the broader context of the field.

4.1. CR Index Validation

The validation of the CR Index is a fundamental step in assessing its effectiveness as a predictor for the density of PS points in the context of infrastructure monitoring. By comparing CR Index values with the distribution and density of PS points derived from the Ground Motion Service Germany (BBD), it was evaluated how well this index reflects the potential for identifying stable scatterers through the PSI technique. The following discussion is based on histograms, density plots, and land use class comparisons, which collectively illustrate the CR Index’s performance.

The histograms for ascending and descending orbit modes from Figure 10a,b reveal a strong relationship between CR Index values and the number of detected PS points. Both histograms show that the majority of PS points are concentrated in a CR Index range between 50 and 90, with a distinct peak near 80. This trend indicates that areas with optimal geometric and land use characteristics, as captured by the CR Index, are associated with a high density of PS points. A Mann–Kendall trend test was conducted to examine the relationship between the CR Index and the density of PS points (PS/ha). The two datasets of the CR Index and the corresponding PS points of the BBD were used for the test. Results show a strong positive trend in the ascending direction, with a Kendall’s Tau of 0.778 and a p-value of 0.00095, indicating a significant increase in PS point density as the CR Index rises. In the descending direction, a similar but slightly weaker positive trend was observed, with Kendall’s Tau at 0.689 and a p-value of 0.0047, confirming statistical significance. These results indicate a clear and meaningful correlation between higher CR index values and higher PS point density. It can, therefore, be concluded that the CR index is a powerful index for the identifiability of PS. Conversely, regions with low CR Index values, typically below 40 (Figure 10c), correspond to a negligible number of PS points. This result aligns with expectations, as unfavorable geometry (e.g., steep slopes with adverse radar incidence angles) and non-reflective land use types (e.g., vegetated areas) tend to hinder the detection of stable scatterers. The similarity in PS point distribution across both ascending and descending modes further highlights the robustness of the CR Index. The consistent performance across different observation geometries suggests that the index effectively accounts for variations in satellite flight paths and radar angles.

The density plot in Figure 11, which examines the number of PS points per hectare (PS/ha) within specific CR Index intervals, provides additional insights into the predictive capability of the CR Index. The results show a clear and consistent increase in PS density with higher CR Index values. The PS density is negligible for CR Index values below 40, begins to rise steadily between 40 and 60, and reaches its maximum in the 80 to 100 range. This trend confirms that the CR Index effectively captures the progressive improvement in PS detectability as geometric and land use conditions become more favorable. Interestingly, there is a slight decline in PS density beyond a CR Index value of 90. This may be attributed to data saturation effects, where certain conditions lead to redundancy in the number of PS points, or limitations in the resolution of the input datasets. Despite this minor discrepancy, the overall trend strongly supports the validity of the CR Index.

The comparison of PS point density across land use classes adds another layer of validation for the CR Index. It is shown for all land use classes in Figure 11. Urban and built-up areas, such as residential zones, industrial regions, and transportation infrastructure, exhibit the highest PS densities. This finding aligns with the high CR Index values typically associated with these areas, as they are rich in artificial structures that serve as stable scatterers for radar signals. For example, residential areas show consistently high PS densities, making them ideal targets for PSI-based monitoring. In contrast, vegetated and water-dominated areas, such as forests, swamps, and moors, demonstrate significantly lower PS densities. These land use types are characterized by dynamic surface properties (e.g., vegetation movement, water absorption), which reduce their suitability for PSI and result in correspondingly low CR Index values. Mixed-use areas and regions with special functional characteristics show moderate PS densities, reflecting their diverse surface compositions and variable potential for radar backscatter.

Notably, the ascending mode consistently produces higher PS densities than the descending mode across most land use classes. This discrepancy may be attributed to differences in radar geometry, particularly in how the satellite’s incidence angle interacts with specific surface features. The results underscore the importance of analyzing both ascending and descending datasets to obtain a comprehensive understanding of PS point distributions. The results of this validation study strongly affirm the utility of the CR Index as a reliable predictor of PS density. The observed correlation between high CR Index values and increased PS densities confirms the underlying methodology, which combines geometric factors (captured in the R Index) with land use data (represented by the Land Use Index, LUI) and not only considers one of the two. The index effectively prioritizes areas with favorable conditions for PSI, providing valuable guidance for selecting regions for infrastructure monitoring. The influence of land use characteristics, as demonstrated by the density variations across land use classes, highlights the importance of incorporating detailed land cover information into the CR Index. By doing so, the index ensures that areas with inherently low PS potential, such as vegetated or water-dominated regions, are accurately flagged as unsuitable for PSI. This capability significantly enhances the index’s practical applicability.

Despite its overall success, minor discrepancies at the extreme ends of the CR Index scale suggest potential areas for refinement. For instance, the slight drop in PS density beyond CR Index values of 90 indicates that further adjustments to the index’s weighting or resolution may be beneficial. Future studies should explore these edge cases in greater detail, leveraging additional datasets and environmental conditions to improve the index’s accuracy.

The validated CR Index offers significant practical benefits for PSI applications, particularly in the context of infrastructure monitoring. By providing a reliable metric for assessing PS potential, the index enables efficient allocation of monitoring resources. For example, areas with high CR Index values can be prioritized for detailed analysis, while regions with low values can be excluded from further consideration. Additionally, the integration of ECRs in areas with moderate CR Index values can further enhance observation capabilities, as these reflectors improve the detectability of PS points. In summary, the validation results demonstrate that the CR Index is a robust and effective tool for predicting PS density. Its integration of geometric and land use parameters provides a comprehensive framework for identifying areas suitable for PSI applications. The consistency of its performance across ascending and descending modes, as well as its strong correlation with PS densities across CR Index ranges and land use classes, underscores its utility for operational infrastructure monitoring. Future work should focus on refining the index for edge cases and exploring its applicability in diverse geographic and environmental contexts to maximize its impact.

4.2. ECR Operation

The variability in Sigma-0 values across sites highlights the role of environmental and geometric factors. Local conditions, such as vegetation, surface roughness, and moisture levels, significantly influence radar backscatter. Additionally, the directional differences between ascending and descending modes suggest that radar incidence angles and reflector orientation are critical for stable signal returns. The stability of the values can be assessed by evaluating the standard deviation in relation to the mean. A standard deviation that is less than 10% of the mean indicates low variation, meaning the values are highly stable and consistent. If the standard deviation falls between 10% and 30% of the mean, the variation is moderate, suggesting a balanced distribution with some expected fluctuations. When the standard deviation exceeds 30% of the mean, the variation is considered high, indicating significant dispersion and lower reliability in the dataset.

Applying this classification to the analysis, the ascending dataset generally exhibits low to moderate variation, demonstrating a higher degree of stability. In contrast, the descending dataset shows greater variability, particularly in certain columns where the standard deviation exceeds 30% of the mean. This increased dispersion suggests that the descending values are less stable and may require additional adjustments or considerations in analytical applications.

The Amplitude Dispersion Index (ADI) serves as a key metric for evaluating the suitability of structures for PSI analysis and can be classified into three main stability categories. ADI values below 0.25 indicate excellent stability, where the scatterers are highly reliable, making these areas ideal for PSI analysis with minimal challenges [53]. Values ranging up to 0.4 represent moderate stability for PSI analysis. In these regions, PSI analysis is still feasible, but certain limitations may arise due to occasional instability of scatterers. Additional measures, such as the incorporation of ECRs, could enhance the reliability of observations in such areas. Conversely, ADI values exceeding 0.4 signify low stability, where scatterer variability is too high for effective PSI analysis. Without external interventions, these areas are considered unsuitable for reliable deformation monitoring. This classification provides a practical framework for assessing and improving the observability of structures in radar interferometric studies [53,54].

In the ascending flight direction, Bigge Reservoir (Bigge 1 and Bigge 2) shows excellent stability, with ADI values of 0.12 and 0.07, respectively. These values fall well below the threshold of 0.25, indicating highly stable conditions suitable for PSI analysis. Both the land- and water-facing sides demonstrate strong scatterer reliability, making additional measures unnecessary. Similarly, the Lister Reservoir achieves an ADI of 0.06, confirming its exceptional stability. This reservoir’s minimal vegetation coverage and optimal orientation enhance its observability, further supporting its inclusion as a prime candidate for PSI analysis. Conversely, the Sorpe Reservoir exhibits an ADI value of 0.52, which is significantly higher than the acceptable range for reliable PSI applications. This instability is attributed to the extensive vegetation covering the embankment, which limits the presence of natural scatterers. The high ADI highlights the need for intervention, such as the installation of ECRs, to improve stability and enhance PSI applicability.

In descending geometry, Bigge Reservoir maintains moderate stability, with ADI values of 0.34 and 0.36 for Bigge 1 and Bigge 2, respectively. These values remain within the range of 0.25–0.4 and they suggest some instability. Nonetheless, the reservoir remains a viable candidate for PSI analysis. The Lister Reservoir, however, exhibits an ADI of 0.92 in the descending flight direction, indicating significant instability. This high value limits its usability for PSI without external stabilization measures, such as ECRs. The discrepancy between ascending and descending geometries highlights the influence of acquisition geometry or geometrical characteristics of the dam wall, or double bounce effects due to different water heights, on ADI values. This last point is subject to future investigations. The Sorpe Reservoir shows notable improvement in descending geometry, with an ADI of 0.12. This value indicates good stability, particularly on the water-facing side, making it a suitable candidate for PSI analysis in this flight direction. The descending flight direction appears to mitigate some of the challenges posed by vegetation and topographical features in the ascending geometry. Overall, the ECRs can be classified as stable enough for a PSI analysis, but it remains to be seen how the stability changes with increasing runtime and whether there are significant differences between individual devices.

Device operation through WiFi and Power supply remains a challenge. Power cables are often not designed for outdoor use and solar panels depend heavily on weather conditions. Environmental factors also impact ECR functionality. Water ingress poses a significant risk to device integrity, while snow coverage can obstruct proper operation. Passive corner reflectors may be too big for dams and because there are no commercial alternatives, ECRs, which are challenging but possible to operate, are the only alternative option to passive CRs. When using ECRs, the occurrence of side lobes must be taken into account, which is why they should have a minimum separation of approx. 200 m if several devices are used at one location. Additionally, individual ECRs have shown varying performance. For example, while the ECRs at Bigge dam and Lister have delivered consistent results, others, such as those at Möhne dam and Verse dam, have faced power-related issues, including inactive batteries or battery protection mode activation. However, as can be seen from Figure 9, the use of ECR results in significantly higher backscatter values compared to times when no measurement took place.

Due to their high cost, ECRs are deployed selectively rather than at a fixed density per square kilometer. Unlike passive reflectors, which can be used extensively, ECRs are strategically placed at specific dam structures where PSI would otherwise be ineffective due to poor visibility, geometric constraints, or low natural scatterer density. Their placement is guided by topographic conditions, SAR visibility, and PSI-based assessments, particularly in areas with high ADI values, where natural scatterers are insufficient. This targeted deployment minimizes costs while maximizing PSI applicability, ensuring effective monitoring of critical infrastructure without requiring widespread installation.

5. Conclusions

This study highlights the critical role of electronic corner reflectors in improving the applicability of persistent scatterer interferometry for monitoring dam and reservoir infrastructure. By integrating the CR Index—a metric combining geometric and land use parameters—optimal installation sites were identified to enhance radar observability. The results validate the CR Index as an effective predictive tool, ensuring the strategic deployment of reflectors to overcome challenges posed by low natural scatterer density and complex topographical conditions. Ultimately, the implementation of electronic corner reflectors enables interferometric analysis in areas where it would otherwise be infeasible, significantly enhancing the potential for long-term infrastructure monitoring.

The operational assessment of the installed reflectors revealed both strengths and challenges. While those at Bigge and Lister exhibited stable radar backscatter signals, others encountered issues related to power supply, environmental exposure, and alignment, particularly in descending radar geometries. The ADI suggest that PSI analysis using ECRs is possible. Despite these challenges, electronic reflectors offer significant advantages over traditional passive reflectors due to their compact size and adaptability.

The combination of persistent scatterer interferometry and electronic reflectors is increasingly recognized as a transformative approach to critical infrastructure monitoring. By addressing limitations such as low scatterer density in complex geometries, reflectors significantly expand the applicability of interferometric techniques to a wider range of structures, making reliable deformation analysis possible, even in challenging environments. Future research should focus on integrating these techniques with emerging high-resolution radar satellites (X-Band) to further refine the precision and coverage of deformation monitoring systems. This integrated approach has the potential to enhance the safety and operational longevity of dams and bridges worldwide.

Looking ahead, the synergy of electronic reflectors, interferometry, and CR Index-based site selection lays the groundwork for scalable, high-precision monitoring strategies for critical infrastructure. Key research directions include refining the CR Index for diverse geographic and environmental conditions, improving hardware to address weather dependence and vandalism, and incorporating multi-sensor data for more comprehensive monitoring solutions. Additionally, advancements in automated maintenance and signal validation will be essential for ensuring long-term operational efficiency.

By tackling these challenges, this study contributes to the advancement of radar-based infrastructure monitoring, strengthening the safety, resilience, and sustainability of dams and reservoirs amid growing environmental and operational demands.