Anne-Sophie Mreyen

SUMMARY: Surface waves such as Rayleigh, Love and Scholte waves can exhibit dispersion, that is, variations in phase velocity with wavelength as a function of frequency. This property enables the inversion of 1-D models of seismic velocity and density in the subsurface. Conventional deterministic and stochastic inversion schemes are widely applied to surface wave data but face two main challenges. The first is the identification of dispersion curves for fundamental and higher modes on wavefield-transformed images, which is often done manually. The second is the quantification of uncertainty, which can be computationally expensive in stochastic approaches or limited to data-propagated uncertainty in deterministic inversions. Our objectives are to (1) eliminate the need for manual or automatic dispersion curve picking, and (2) directly infer ensembles of 1-D velocity models—and their associated uncertainties—from the full velocity spectrum, that is, the complete dispersion image containing all modes. To this end, we employ Bayesian Evidential Learning, a predictive framework that reproduces experimental data from prior information while allowing prior falsification. In our application, ensembles of prior Earth models are sampled to predict 1-D subsurface structures in terms of seismic velocity and, where applicable, attenuation from near-surface seismic wave data. This approach bypasses traditional inversion schemes and provides a computationally efficient tool for uncertainty quantification.

The Pays de Herve, located in the eastern part of Belgium, can be characterized as a multiple sections tableland with gentle slopes of less than 15°. It is located in the vicinity of the northern section of the Hockai Fault Zone, a 42 km-long seismogenic fault zone, that is characterized by the presence of fault scarps, multiple dissection elements and the presence of more than 20 paleo-landslides. Among these latter, the Manaihan landslide is the most studied and monitored landslide in the area. From a geological point of view, it developed in a Upper Cretaceous sedimentary setting, i.e., Vaals Clays overlaying Aachen sands. Even today, the slope is affected by instability and subsidence phenomena, likely linked to anthropogenic loading combined with prolonged periods of rainfall and possibly historic seismic events leading to liquefaction in the Aachen sands.Recently, new geophysical surveys have been carried out using an integrated approach, combining electrical resistivity measurements, active seismic methods (interpreted as P-wave tomography and MASW), and passive seismic techniques (single-station H/V). The key question to address is: How deep is the sliding surface, and is it possible to identify it?The combination of these surveys allowed the identification of two main layers. The first layer has a variable thickness ranging between 5 m and 20 m. On the electrical resistivity tomography, it corresponds to a more conductive layer with values between 5 and 20 Ωm, and on the seismic tomography, it shows velocities between 500 and 1300 m/s. From the electrical and seismic tomographies, the second layer appears to be more resistive, with values ranging between 30 and 50 Ωm, and P-wave velocities exceed 1500 m/s. Based on the geological map and their physical properties, the identified layers have been attributed to the Vaals clay formation and the Aachen sands, respectively.The H/V measurements were processed to produce sections showing the variation of H/V amplitude (or the log of H/V) with depth. If multiple H/V measurements can be aligned in a linear array and the surface layer can be assumed to be homogeneous, i.e., shear wave velocity increasing with depth, the H/V curves along the alignment can be modeled together to create a 2D section. Several H/V sections could be developed for the Manaihan landslide, revealing a similar pattern of contrasts between the before mentioned layers. The main contrast is located at a depth ranging from 15 m to 40 m and corresponds to the interfaces identified by ERT and SRT. This interface is present beyond the landslide and even outside of it, suggesting that it may be associated to the geological contact between the Vaals clays and the Aachen sands. The conductive layer identified in the ERTs can furthermore be associated to very low log H/V amplitudes in the upper range of the H/V sections until 10-20 m depth. The H/V amplitude analysis of all the identified sections suggests that the sliding surface of the Manaihan landslide is located at the contact between the clay and the sand layers.

While geodetic measurements have long been used to assess landslides, seismic methods are increasingly recognized as valuable tools for providing additional insights into subsurface structures and mechanisms. This work aims to characterize the subsurface structures of the deep-seated gravitational slope deformation (DSGSD) at Heinzenberg through the integration of active and passive seismic measurements. Seismic techniques can hereby deliver additional information on the subsurface structure and mechanisms involved, e.g., the degree of rock mass degradation, the resonant frequencies of the potentially unstable compartments, and the local fracture network orientations that are influenced by wavefield polarization. By employing advanced methods such as H/V analysis, site-to-reference spectral ratios, polarization analysis, surface wave analysis, and the joint multizonal transdimensional Bayesian inversion of velocity structures, we establish a comprehensive baseline model of the landslide at five selected sites. This baseline model shall help identify potential changes after the refilling of Lake Lüsch, which started in 2021. Our results reveal the rupture surface of the DSGSD at various depths ranging from 30 m at the top to over 90 m in the middle of the slope. Additionally, we estimate key parameters including the shear wave velocities of the different rock masses. The 2D geophysical profiles and rock mass properties contribute to the understanding of the subsurface geometry, geomechanical properties, and potential water pathways. This study demonstrates the significance of integrating seismic methods with traditional geodetic measurements and geomorphologic analysis techniques for a comprehensive assessment of landslides, enhancing our ability to monitor and mitigate hazardous events.

Prior geological knowledge is crucial in a wide range of geophysical case studies aiming at data processing and interpretation. In this regard, geological modeling methods can be considered as efficient tools to facilitate petrophysical parametrization and constrain geophysical inversions as a priori information over a defined model space. We use the open-source library GemPy (De la Varga et al., 2019) which offers a community driven alternative based on the potential-field method and specialized in probabilistic modeling.In this work, we show an exemplary case study of high-resolution marine geophysical data comprising single-trace reflection and underwater refraction seismics validated with geotechnical data. Two approaches are suggested: (1) digitalization of geophysical imagery, i.e., spatial information from interpreted horizons and inherent uncertainties, and (2) pseudo depth migration of picked reflection seismic travel times using a simple velocity model from parallel recorded underwater refraction data. Next to a deterministic “best-fit” solution, the model is interpreted following probabilistic distributions of input data and classified after their identified certitudes (e.g., depth range of an observed seismic horizon), where prior knowledge is optionally included using Bayesian Networks. Finally, global uncertainties are estimated by multiple model realizations allowing for improved data assessment and enhanced decision making.The further outlook of this study is the creation of a variety of digital twins taking into account realistic conditions in terms of both, the geological environment as well as data acquisition, as a solid prior for future data exploitation by inverse processes of wave propagation in shallow marine environments.

The E-TEST project (Einstein Telescope EMR Site & Technology) investigates the feasibility of constructing a large Laser Interferometer (Einstein Telescope) in the Euregio Rhine Meuse. The aim of this instrument is to detect gravitational waves. To reach a sufficient noise attenuation the telescope will be built deep underground (more than 200 meters depth). The infrastructure consists of multiple tunnels and caverns spanning several kilometers. As with any large-scale infrastructure, the geological model, especially the existence and orientation of faults, is of large importance for hydrogeophysical and geotechnical characterization. At such depths, few near-surface geophysical methods are able to provide information with enough details. The application of large 3D Deep ERT surveys helps understanding the local geological settings and to identify important geological features but suffers from ambiguous interpretation. However, such imaging requires measuring dipoles independent of the injection system (such as the Fullwaver System by IRIS Instruments) in contrast to conventional ERT Systems, and due to the large covered area, coarsely spaced. This results in a drastic decrease in resolution when compared to classical ERT measurements.Here, we present a sensitivity analysis on a dataset based on the geologic setting in Plombières, Belgium to identify the impact of geology and survey setup on Deep 3D ERT surveys. The utilized geologic model was created with the Open-Source 3D structural geomodelling software GemPy and used as an input for forward modelling using the Open-Source modelling and inversion library pyGIMLi.

The San Andrés landslide on El Hierro (Canary Islands) represents a rare opportunity to study an incipient volcanic island flank collapse with an extensive onshore part. The presented research improves the knowledge of the internal structure and rock characteristics of a mega-landslide before its complete failure. The investigation combines multiple geophysical measurement techniques (active and passive seismic) and remotely sensed, high spatial resolution surveys (unmanned aerial vehicle) with in situ and laboratory geotechnical descriptions to characterize the rock properties inside and outside the San Andrés landslide. The available geophysical and geological data have been integrated into 3D geomodels to enhance their visual interpretation. The onshore geophysical investigations helped detect the possible San Andrés landslide sliding surfaces at depths between 320 m and 420 m, with a rather planar geometry. They also revealed that rocks inside and outside of the landslide had similar properties, which suggests that the previous fast movements of the landslide did not affect the bulk properties of the displaced rocks as the failure chiefly occurred along the weakened sliding plane. Uniaxial strength tests on basalt rocks further indicate a high variability and spatial heterogeneity of the rock strength properties due to the different types of volcanic rocks and their texture. The new information on the rock properties and structural setting of the San Andrés landslide can now be used to develop realistic geotechnical slope models of the onshore part of the flank collapse that are possibly applicable for slope stability or deformation calculations. It will also help assess related hazards marked by a low occurrence probability and a high impact potential.

First analyses of landslide distribution and triggering factors are presented for the region affected by the 14 August 2021 earthquake (M_w=7.2) in the Nippes Department, Haiti. Landslide mapping was mainly carried out by comparing pre- and post-event remote imagery (∼0.5–1 m resolution) available on Google Earth Pro^® and Sentinel-2 (10 m resolution) satellite images. The first covered about 50 % of the affected region (for post-event imagery and before completion of the map in January 2022), and the latter were selected to cover the entire potentially affected zone. On the basis of the completed landslide inventory, comparisons are made with catalogs compiled by others both for the August 2021 and the January 2010 seismic events, including one open inventory (by the United States Geological Survey) that was also used for further statistical analyses. Additionally, we studied the pre-2021 earthquake slope stability conditions. These comparisons show that the total number of landslides mapped for the 2021 earthquake (7091) is larger than the one recently published by another research team for the same event but slightly smaller than the number of landslides mapped by a third research team. It is also clearly smaller than the one observed by two other research teams for the 2010 earthquake (e.g., 23 567, for the open inventory). However, these apparently fewer landslides triggered in 2021 cover much wider areas of slopes (>80 km²) than those induced by the 2010 event (∼25 km² – considering the open inventory). A simple statistical analysis indicates that the lower number of 2021 landslides can be explained by the missing detection of the smallest landslides triggered in 2021, partly due to the lower-resolution imagery available for most of the areas affected by the recent earthquake; this is also confirmed by an inventory completeness analysis based on size–frequency statistics. The much larger total area of landslides triggered in 2021, compared to the 2010 earthquake, can be related to different physical reasons: (a) the larger earthquake magnitude in 2021, (b) the more central location of the fault segment that ruptured in 2021 with respect to coastal zones, (c) and possible climatic preconditioning of slope instability in the 2021 affected area. These observations are supported by (1) a new pre-2021 earthquake landslide map; (2) rainfall distribution maps presented for different periods (including October 2016 – when Hurricane Matthew had crossed the western part of Haiti), covering both the 2010 and 2021 affected zones; and (3) shaking intensity prediction maps.

First analyses of landslide distribution and triggering factors are presented for the region affected by the August, 14, 2021, earthquake (Mw=7.2) in the Nippes Department, Haiti. Landslide mapping was mainly carried out by comparing pre- and post-event remote imagery (~0.5–1-m resolution) available on Google Earth Pro® and Sentinel-2 (10-m resolution) satellite images. The first covered about 50 % of the affected region (for post-event imagery and before completion of the map in January 2022), the latter were selected to cover the entire potentially affected zone. On the basis of the completed landslide inventory, comparisons are made with catalogues compiled by others both for the August 2021 and the January 2010 seismic events, including one open inventory (by the United States Geological Survey) that was also used for further statistical analyses. Additionally, we studied the pre-2021 earthquake slope stability conditions. These comparisons show that the total number of landslides mapped for the 2021 earthquake (7091) is larger than the one recently published by another research team for the same event, but it is also clearly smaller than the one observed by two other research teams for the 2010 earthquake (e.g., 23,567, for the open inventory). However, these apparently fewer landslides triggered in 2021 cover much wider areas of slopes (>80 km²) than those induced by the 2010 event (~25 km² – considering the open inventory). A simple statistical analysis indicates that the lower number of 2021-landslides can be explained by the missing detection of the smallest landslides triggered in 2021, partly due to the lower resolution imagery available for most of the areas affected by the recent earthquake; this is also confirmed by an inventory completeness analysis based on size-frequency statistics. The much larger total area of landslides triggered in 2021, compared to the 2010 earthquake, can be related to different physical reasons: a) the larger earthquake magnitude in 2021; b) the more central location of the fault segment that ruptured in 2021 with respect to coastal zones; c) and possible climatic preconditioning of slope instability in the 2021-affected area. These observations are supported by (1) a new pre-2021 earthquake landslide map, (2) rainfall distribution maps presented for different periods (including October 2016 – when Hurricane Matthew had crossed the western part of Haiti), covering both the 2010 and 2021 affected zones, as well as (3) shaking intensity prediction maps.

The present study aims at the investigation of the large (> 10 km run-out and volume up to 20-50 km3), ancient (176 and 545 ka) coastal and submarine San Andres landslide on volcanic island, El Hierro. The landslide formed, as a result of an aborted giant volcanic flank collapse and represents a rare site where the landslide mass and related failure planes can be studied onshore. The possible triggers of a future massive failure include volcanic activity, large earthquakes, and the rising sea level under future climate change scenarios. Therefore, it is important to analyze possible associated geological hazards based on an extensive site study, as a massive failure could have catastrophic consequences, not only for the population of El Hierro Island, but also for the neighboring islands and even beyond. To that end, we adopted an integrated geophysical approach including horizontal-to-vertical spectral ratio (H/V), seismic array measurements, multi- channel analysis of surface waves (MASW) and seismic refraction tomography (SRT). The data acquisition phase included campaigns in 2020 (H/V and seismological arrays) and 2021 (H/V, MASW, SRT). The seismic survey (refraction and MASW) was completed with profiles of variable lengths using 48 geophones while ambient noise array and single station measurements were carried out with CMG-6TD broadband velocimeters. Also a mixed scheme combining geophones and seismic stations along longer profiles were tested to increase the investigation depth. Data collected along these profiles were processed both in terms of SRT (providing a 2D P- wave velocity distribution over varying depths) and of MASW (providing S-wave velocity logs through surface wave inversion). Such S-wave velocity logs were also computed by inverting the array seismic noise data (first processed by f-k and cross-correlation techniques, providing a surface wave dispersion curve), which generally provided related information over larger depths. Denoising of the seismic refraction data was often necessary before picking of the first arrivals as the records were affected by a high level of noise due to the proximity of coastal areas and of windmills. Our approach increased the precision in picking the first arrival and in determining the depth of the elastic properties of the landslide. Additionally, a geomodeling approach is used for the better presentation of the results in 3D. The latter reveals large variations in the measured physical properties because of the highly heterogeneous conditions marking the volcanic environment. The delineated subsurface information will provide an essential input parameter to be used for further numerical modeling studies of the flank stability and of the potential impact of a collapsing mass on the ocean.