Tectonic landforms in northeastern Belgium, located in the most seismically active region of stable Europe (Figures 1A, B) northwest of the Alps, are generally modest in stable plate interiors. However, it appears that the major faults marking the western edge of the Roer Valley Graben (RVG)—the Feldbiss and Heerlerheide faults—may extend toward the southeast into the central part of the Stavelot massif as far as the High Fens (Hautes Fagnes in French) ridge (Demoulin, 1988). The graben is bounded by the Peelrand fault to the northeast and the Feldbiss fault zone to the southwest, which are two active normal fault systems trending NNW–SSE (Camelbeeck et al., 2007). Until now, research conducted in the RVG has mainly been based on geophysical investigations, consisting primarily of electrical prospecting and tomography, seismic refraction, and ground-penetrating radar (GPR) (Meghraoui et al., 2000). An additional method that may be used in geophysical prospecting is high-resolution seismic reflection profiles (Demanet et al., 2000). These techniques, in combination with geological and geomorphological investigations, enable approximate positioning of the faults. In addition, remote sensing performed by ground scanning lasers provides additional information about slight ground movements, which are not otherwise detectable.

Synthetic aperture radar (SAR) interferometry (InSAR) has been widely used to analyze Earth surface displacement processes such as landslides, subsidence, and volcanic and seismic phenomena (Gabriel et al., 1989). Modern-generation SAR satellites obtain a spatial resolution of several meters (e.g., Sentinel-1) with a revisit time of some days down to 24 h (e.g., Iceye), allowing the increasingly more reliable detection of differential displacements which can be used for monitoring and assessing the potential critical behavior of single structures or infrastructures (Talledo et al., 2022). Persistent scatterer interferometry (PSI) (Crosetto et al., 2016) and small baseline subset (SBAS) (Li et al., 2022) techniques allow us to estimate the projection of displacements in the line-of-sight direction—that is, the displacement component away or towards the sensor (according to ascending and descending orbits). High-precision (less than 1 cm) deformation monitoring and high accuracy in target georeferencing can be achieved using PSI (Ullo et al., 2019). Devleeschouwer et al. (2008) compared PSI to ENVISAT data between 2003 and 2010 with gravimetry measurements to reveal an uplift over NE Belgium. Recently, the European Gound Motion Service (EGMS) Ortho has provided harmonized information regarding natural and anthropogenic vertical and east–west ground motion phenomena resampled with a 100-m spatial resolution over Europe (Crosetto et al., 2020).

A less populated area in the western part of the RVG has been selected as a candidate to accommodate, at a depth of 250 m, the future triangular Einstein Telescope (ET, with side lengths of 10 km) which will observe gravitational waves. Before an international jury decision in 2025, many criteria must be considered when selecting the site, including geology, hydrogeology, societal aspects, and the lifetime and cost of the ET. Geological conditions are difficult to establish over the large scale of the telescope. Three transverse fault zones are observed in the north of the Pays de Herve (BE), around the potential ET area (Figure 1C): 1) the Eupen faulting zone stretching from the Gueule Valley in the Pays de Herve to the flanks of the Pannensterzkopf on the High Fens crest; 2) the Baelen faulting zone stretching from the Gulp valley in the NW to the western slope of the Botrange peak of Belgium in the southeast; 3) the Hockai fault zone (HFZ), which forms the western boundary of the vast network of transverse Neogene and Quaternary faulting activity (Demoulin, 1988; Camelbeeck et al., 2000). The HFZ in the south of the ET region produced the largest known earthquake in Northwest Europe: the 1692 Verviers earthquake with a magnitude of about 6.3 (Camelbeeck et al., 2000). The NW–SE trending normal–transverse faults could be seismically active while the NE–SW trending thrust faults are old (Variscan) inactive structures. Kreemer et al. (2020) suggested that uplift observed in this region could be related to a buoyant mantle plume under the Eifel. The latter could indeed be responsible for the regional ground uplift, including the Weser–Geul (BE) and South Limburg regions (NL), and thus could also affect the area selected for the future ET installation.

The activity of the HFZ is also marked by the presence of numerous slow-moving landslides in the Battice (Herve) region (Demoulin and Collignon, 2000; Demoulin et al., 2003; Demoulin and Glade, 2004; Demoulin and Chung, 2007) and by a series of aligned but not continuous NW–SE oriented scarps (related to the aforementioned transverse faults) that can be found between the Battice and Malmedy regions—the most pronounced being the 150-m-long and 20-m high Beverce scarp adjacent to an old massive landslide (Mreyen et al., 2018). The seismic activity of faults in the eastern ET area is not well-known, but widespread karst phenomena and the “Calamine-type” (Pb-Zn) mineralization of faults in the Gueule Valley (between La Calamine and Plombières) at least hint at the presence of numerous tectonic structures (Dejonghe and Boni, 2005). North of the area of interest (AOI), fault reactivation induced in 1936 by coal mining caused a 0.34–0.4-m vertical displacement of the base of the plough layer at the Heerlerheide fault, part of the southern boundary fault system of the Roer graben—the Feldbiss fault zone (Van Balen et al., 2021).

This study performed a detailed analysis along the numerous normal (and, partly, also strike-slip) faults crossing the ET area, focused on the Herve–Aubel–Voeren–Geul region in Belgium and in South Limburg, the Netherlands. The main hypothesis assumes that most of these faults are not considered active, which has been confirmed by remote sensing (InSAR techniques) and ground-based observation before the implementation of integrated geodetic reference stations (IGRSs). An IGRS consists of a GNSS receiver and two InSAR corner reflectors (for ascending and descending tracks) installed on the same deeply founded monument (Ketelaar et al., 2020). The advantage of IGRSs is that the accuracy of leveling, GNSS, and InSAR deformation estimates can be directly cross-validated because the measurements refer to the same monument and therefore reflect the same cause of deformation. This scientific contribution also aims to answer these principal research questions. Can both PSI and SBAS techniques differentiate local and regional ground uplift and sub-regional subsidence, possibly enhanced across fault structures? Do old mine activities also influence the observed displacement patterns? According to Vervoort and Declercq (2018), local features such as geological faults can have a secondary effect on the local variation of the uplift. The processes of subsidence and uplift are based on completely different mechanisms: are faults observed in AOI aseismic or SAR sensor resolutions not able to detect such small displacements? In comparison, clearly visible neotectonic deformation will be analyzed and inferences will be made according to the detectability of Earth surface displacements according to their possible origin(s).

This research focused on determining surface deformation using SAR interferometry and geomodeling in the transborder Meuse–Rhine region. The Figure 2 shows the individual stages of workflow. The next subsections describe in detail the data used and their processing, enabling the development of 3D visualization.

This research conducted a detailed analysis along the numerous normal (and partly also strike-slip) faults crossing the ET area focused on the Herve–Aubel–Voeren–Geul region in Belgium and in South Limburg, the Netherlands. We focused on the following areas: the Gueule and Gulp faults in the Pays de Herve (BE), the Heerlerheide fault (NL) in the RVG, the railway north of Remersdaal (BE), and an active landslide zone in the Battice region (BE). In all areas, ground deformations were studied based on the results of the InSAR technique—particularly the PSI, MSBAS, and PSBAS methods.

Significant positive regional deformations were observed in the RVG area, where the displacement velocities were up to 10 mm/yr. This deformation signal may be related to the numerous geological faults in this area. Cuenca et al. (2013) used SAR data—ERS-1, ERS-2, and Envisat—and the PSI technique and observed most of the graben uplifts with respect to adjacent horsts at velocities of 1 mm/yr in the Roer graben (1992–2010). In addition, they linked the appearance of a larger uplift signal of approximately +4 mm/year in space and time with the cessation of pumping in the Erkelenz coalfield, adjacent to the RVG. Camelbeeck and Meghraoui (1998) concluded that tectonic activity in the RVG appears relatively stable, at least in the last 150 kyr. Within the regional scale deformation, we detected maximum negative LoS displacements in the open-pit area of the Hambach mine of about −40 mm/yr. In comparison, Tang et al. (2020) detected subsidence greater than −500 mm/yr for 1 year (2017/2018) over reclaimed areas based on the SBAS technique and Sentinel-1 and TerraSAR-X images. As the main factor of subsidence, they indicated a process of compaction of waste materials. In the ET triangle, we did not observe significant displacement velocities (i.e., around 2 mm/yr) and the values were similar in all InSAR results.

Analysis of InSAR results along the Gulp fault (which could split in two near the village of Hombourg) showed that points in the north are uplifting while points in the south are subsiding. This was also confirmed by the results obtained in the calculation of the double difference made for the benchmarks Hb1 and Hb2, whereas the highest positive values were observed to the west of the western fault branch, possibly indicating a foot wall (Figure 4). Annual velocity rates of the benchmarks extracted along the Gueule and Gulp faults were less than −2 mm/yr—an insignificant value. However, by comparing the velocity values for the extracted benchmarks along the faults, it can be seen that the Gulp fault (Hb1 and Hb2) near Hombourg is characterized by slightly higher annual velocities than the Gueule fault (Ge1, Ge2, and Ge5) (Table 4).

A railway line passes through the Gulp fault, from which we extracted two points for analysis (RW1 and RW2). The maximum deformation rates were −3.8 mm/yr in the RW1 benchmark, which subsided faster than the RW2 benchmark (Table 4). Our time series analyses based on InSAR results along the Heerlerheide fault indicated that its eastern face is uplifting at a velocity of up to 8 mm/yr (Table 4). In the case of the western wall, the main LoS displacement trend was not identified. Based on the MSBAS results, we concluded that the displacement rates decrease towards the south. According to Van Balen et al. (2021), the vertical coseismic displacement of the youngest event was around ∼0.25 m (17–15 ka), while the estimated moment magnitude is around 6.2. Additionally, the Heerlerheide fault experienced a mining-induced vertical offset of about 0.4 m in 1936 (Heitfeld et al., 2016). In interpreting the results for the Heerlerheide fault, it should be emphasized that, from the end of the 19th century to the first half of the 20th century, coal was mined south of the Feldbiss fault zone. Van Balen et al. (2021) indicated that the upper part of the fault was reactivated by deep exploitation. This phenomenon highlights the fact that a fault is a zone of structural weakness. Moreover, according to Pöttgens (1985), Caro Cuenca et al. (2013), and Vervoort (2016), re-flooding of the mines after their closure caused a regional rebound that continues to the present. The InSAR time series analysis confirmed the presence of uplifts in the studied region. This rebound counters only a fraction of the mining which had induced subsidence, is gradual, and is not confined to the former mining concessions (Heitfeld et al., 2016). It is worth noting that the fault has not been reactivated during the re-flooding of the mines and the resulting surface rebound; however, this could happen because the rebound is still ongoing. Considering individual benchmarks along the faults, we noticed that the Heerlerheide fault is characterized by higher velocity rates than the Gueule and Hombourg faults when considering the MSBAS method (Table 4). However, the DepSI method shows the calculated velocities at a similar level. The InSAR results obtained are very small and could be described as insignificant; therefore, we cannot find increased seismic activity in the analyzed faults, especially Heerlerheide and Gueule as old mining activity may be responsible for the observed deformation.

In the Battice area, displacements of less than 3 mm/yr were observed based on InSAR results for 2016–2022. According to Preuth et al. (2010), some dormant landslides in this region have been reactivated by anthropogenic activities such as highway construction, suburban development, or construction of industrial sewer pipes. Unfortunately, the inability of InSAR to analyze ground deformation hazard may be explained by two reasons. First, it could be due to too-high movement rates, which the InSAR technique cannot detect because visible steady-state displacements are limited until 2.8 cm—half the wavelength of C-band radar wave (Yang et al., 2017). The second reason is the vegetation coverage (combined with the absence of “stable” built-up areas) within the landslides, which induces signal temporal decorrelation.

The first reason is documented by the Manaihant landslide example, which proved to be active over recent years, while no meaningful difference in phase values could be obtained. Moreover, the difference between the LIDAR DTMs indicates a continuous uplift in the upper (western) part of the landslide related to a 5 m high (anthropogenic) deposit accumulated over 9 years (2013–2022), corresponding to a displacement rate of around 56 cm/yr. The Manaihant landslide activity had been studied by researchers such as Demoulin et al. (2003), who had also identified earlier partial landslide reactivation with episodic movements since 1980, at a mean rate of ∼20 cm/yr. Demoulin and Glade (2004) found that most landslides have been inactive for many years in the Battice area. Only the Manaihant landslide had been massively reactivated, by heavy rains on 14 September 1998. This included damage to two buildings, allocated close to the newly formed 1.5-m-high scarp above the old head scarp.

InSAR techniques allowed us to detect regional and local ground deformations, revealing trends in the behavior of the LoS velocity rates. During the study period, velocities of approximately 2 mm/yr were observed in the ET Triangle area, which are insignificant due to the measurement uncertainty of the InSAR techniques themselves. Even the faults crossing this area did not show significant displacements, which confirms the initial hypothesis of their low seismotectonic activity. Therefore, we consider the main area of interest as being relatively stable. Nevertheless, the integrated geodetic reference stations (IGRSs) installed in February 2023 in Plombières (province of Liège, Belgium) enabled continuous high-accuracy monitoring of ground deformation to confirm that related signals are anthropogenic. These two stations are located on either side of the Gueule fault—one on the library side (PBLB) and the other on the cemetery side (PBCM). The IGRS was developed and installed within the project “Ground Deformation from Meteorological, Seismic and Anthropogenic Changes Analysed by Remote Sensing, Geomatic Experiments and Extended Reality—GERMANE” (Living Planet Fellowship program). Additionally, constant monitoring here will enable comparison of the IGRS results with other measurement methods such as InSAR. In the case of the RVG–Heerlerheide fault, the detected deformations seem due to the influence of former mining activity, probably combined with seismic activity. It should also be emphasized that, in the Battice area, the main deformation (proven only by comparison between high-resolution DEMs of different recent ages with velocities too large for InSAR methods) is related to the active Manaihant landslide.

Based on our research, we concluded that the main hypothesis regarding the low seismotectonic activity (of recent and present times) of faults located in eastern Belgium has been confirmed using the PSI and SBAS techniques. InSAR techniques do have some limitations: 1-D displacement recorded, computational algorithm and SAR properties, and a lack of results due to the vegetated land cover and in areas where deformation velocities are too fast (e.g., Battice landslide). There are several important aspects to highlight when comparing PSI and SBAS results. PSI and SBAS measure different processes: displacements of coherent points at specific locations vs. spatially distributed deformation, respectively, revealing homogenous patterns that are less sensitive to superficial changes due to ground properties varying through seasons. Permanent (or coherent) scatterers are very rare in the natural environment (due to missing stable natural reflectors); this is why the PSI method is most often used in urban areas in contrast to SBAS. The results of the SBAS method provide a smoother and more detailed displacement time series, leading to better representation of the temporal evolution of ground deformation—also in a natural environment but often less accurate. Nevertheless, it should be emphasized that both methods provide helpful information about ground deformation and can detect small displacements (on a millimeter level) over a relatively long period of time if good coherence values are maintained.

Finally, the prepared virtual environment containing InSAR results and geological data enables a better understanding of the three-dimensional deformation pattern, which may appear very complex in relation to the actual geological context of the area. In addition, our research adds value to the ongoing Interreg E-TEST project (investigating the site for an ET in the area of interest shown here) and may be of use when the final location of the Einstein Telescope is determined.

We formulated answers to the research questions presented at the beginning of the paper.

Can both PSI and SBAS techniques differentiate local and regional ground uplift, and sub-regional subsidence, possibly enhanced across fault structures?

PSI results over long time series of acquisitions have not known any regional ground deformation, only punctual spots (natural or anthropogenic permanent scatters) corresponding to mining. On the other hand, SBAS results give additional information about regional ground deformation observed inside and around the AOI due to the denser coverage of the area with points. We did not observe a significant increase in deformation across fault structures.

Do old mining activities also influence the observed displacement patterns?

No evidence has revealed that old mining activity influences the patterns of the observed displacements. The spatial deformation analysis shows an inhomogeneous subsidence of up to 4 mm/year for the old galleries in the Plombieres region (BE).

Are faults observed in the AOI aseismic or the SAR sensor resolution unable to detect such small displacements?

InSAR results give us good insights for understanding the full deformation picture over the AOI, but limitations in terms of technicality and deformation significance of the displacements are very small.