With an increasing worldwide demand for critical raw materials, among which Rare Earth Elements (REE), mine-waste deposits in Sweden are considered as secondary sources of critical minerals. These so-called mine tailings are abundant, and located close to the surface, facilitating their excavation. To assess the economic potential of these tailings as a source of critical minerals, their volume must be estimated. The horizontal-to-vertical spectral ratio (HVSR) method, which estimates the fundamental frequency of sedimentary sites, is particularly suited for the purpose of mapping the depth-to-bedrock. We apply this method to three-component ambient-noise recordings that were recently collected over the Bäckegruvan mine tailings in the Bergslagen province, central Sweden, to delineate their depth.
Additionally, we employ 3C beamforming to characterize the ambient-noise wavefield, with the aim of supporting HVSR curve interpretation. However, the results of the 3C beamforming prove unreliable, which we attribute primarily to the array's suboptimal geometry - specifically, its sparse station spacing and highly anisotropic wavenumber resolution.
We estimate the thickness of tailings at each station by combining the fundamental site frequency with a constant S-wave velocity. We then interpolate these individual measurements to construct a three-dimensional model of the tailings-bedrock interface, from which we derive the total volume of the tailings deposit. To assess the reliability of our model, we compare the interpolated depths with findings from a recent investigation in the same region, which utilized geoelectrics and electromagnetic geophysical methods. Comparing the results, we find strong concordance between the two approaches.
Our case study highlights the HVSR method’s potential as a cost-effective, fast, and robust approach for obtaining a preliminary estimate of mine-tailings depth.

Fiber-optic distributed acoustic sensing (DAS) excels in high-quality seismic signal acquisition and detection but is often hindered by noise, significantly reducing signal-to-noise ratio (SNR) and impeding microseismic event detection. Moreover, continuous seismic monitoring campaigns generates huge data volumes. While numerous denoising approaches exist, they often demand substantial computational resources, hindering real-time implementation. We propose an unsupervised neural network to suppress random noise without requiring noisefree ground truth or prior noise characteristics. The network learns to extract features by masking random input traces and reconstructing the target using long-term coherence from neighboring traces. We explore hyperparameter optimization by varying input sample generation, activation functions, scaling methods, and the number of input traces. We evaluate the model by running a detection algorithm on FORGE data and achieve a 43% increase in event detection. We further exploit our algorithm in real-time experiments and achieved within a 90% processing time compared to the data acquisition rate with denoising implemented. Our approaches can be incorporated real-time data acquisition, effectively facilitate the screening and storing the data timeframe with useful information.

The Bedretto Laboratory, located within the Rotondo granite, conducts stimulations of the host rock to investigate the possibility of enhanced geothermal energy and tries to better understand the risk of induced seismicity. Seismic anisotropy plays an important role when processing seismic data, making its characterization essential in such a laboratory. In a previous study, through crosshole seismic, the apparent velocity of P-waves was obtained and used to give some first values of the parameters characterising anisotropy assuming a Tilted Transversely Isotropic medium (TTI). Further investigating the obtained data, S-wave splitting could be detected. This phenomenon occurs in anisotropic media, resulting in horizontal S-waves (SH) and vertical S-waves (SV) having different seismic velocities. A sensitivity analysis was performed to assess the significance of S-wave information. The analysis revealed the benefit of incorporating the information provided by the S-waves to improve the current anisotropic model. Several crosshole surveys were carried out using different instruments, such as a S-wave sparker or a 3-component geophone. The aim was to understand the behavior of S-waves better and accurately identify the arrival time of both types of waves. The results reveal that rotating the 3-component geophones data was the best choice to isolate both types of S-waves. Time arrivals were picked, and apparent velocities were calculated knowing the ray path length. They show a change from 2747 m/s to 3216m/s for SH waves, and from 2830m/s to 3105m/s for SV-waves. An attempt was made to obtain the anisotropy parameters that successfully describe the observed velocities in the Rotondo granite.

In Geodesy, the development of accurate Digital Elevation Models of the surface helps to provide height data for various applications in other Earth Science disciplines. These models are used as base models for studies. Geophysics can be a support science when providing base models for non-invasive studies inside the Earth. Such models can be multi-dimensional velocity models, where the relation of velocities and depth can be used for other structural parameter studies in e.g. Geodynamics or Geochemistry. With technological advancements in the recent decade, there is incentive to provide new regional continental-scale models. Full-waveform inversion (FWI) is an imaging method that iteratively inverts through seismic waveforms for structural heterogeneities in the Earth to create a high resolution static im- age. The development of event-specific adaptive SmoothieSEM meshes for regional meshes leads to a substantial decrease in computational costs of the FWI by lowering the amount of mesh elements needed to model a wavefield. This allows to push regional continental-scale tomographies towards period bands that were previously cost-prohibitive. Divided into two tomographic studies, on a bigger continental-scale mesh of Europe and Western Asia and a smaller mesh focused on Europe respectively, the study areas are lowered to a period of 24s. The reference base model is derived from the current generation of the Collaborative Seismic Earth Model (CSEM). Starting at a minimum period of 50s, the mimimum period is pushed in steps towards the target of 24s in a multiscale inversion approach to enable monotonic misfit reduction. The misfit optimization of the synthetic and observed waveforms is based on time-frequency phase misfits while the gradient optimization is based on the trust-region limited memory Broyden-Fletcher-Goldfarb-Shannon (L-BFGS) method and dynamic mini- batches. Waveform fits are quantified by the reduction of the time-frequency phase misfit and normalized amplitude difference misfits between the end model and the CSEM base model. The waveform inversion reveals more heteregeneous structures in Europe and Western Asia compared to previous studies.

Landslides pose a risks to the communities settled in their vicinity. They pose a threat to human lives and significant economical damage to the affected communities. As the scientific community agrees that the climate change will further increase the risks associated with landslides, it is important to develop a reliable, cost effective and widely applicable technique to assess the stability of the potentially unstable slopes. Such a technique could allow to create an early warning systems meant to help evacuate the communities at risk before the landslide event happens. There have been many studies applying different methods of achieving this goal. This study will apply electrical resistivity tomography to investigate the subsurface and use machine learning methods in the form of supervised and unsupervised learning to interpret the measurements. The electrical resistivity tomography is a widely applied technique to investigate the stability of slopes. It can determine the water table in the subsurface, which plays the crucial role in the stability of the slopes. However, the manual interpretation of the electrical resistivity tomography profiles is a time-consuming and cumbersome task. In order to develop a reliable early warning system both the measurements and their interpretation have to be automatized. The first one has been already achieved. There are systems available that can perform electrical resistivity tomography automatically and continuously. In order to automatize the second part, the following research investigated the capability of machine learning to interpret resistivity profiles alongside with various methods of improving the accuracy of the results. The results show that machine learning is capable of performing this task. Some limitations were identified and guidelines for the preprocessing of the data were proposed. Several areas that require further investigation were identified.

Measuring temperature variations of the deep ocean is necessary to evaluate the heat flux between the atmosphere and the hydrosphere and thus calibrate climate change models. In the last two decades a passive alternative to profiling oceanographic floats has emerged: hydroacoustic thermometry. This method consists of using the oceanic ambient noise field as a source of acoustic waves and hydrophone arrays as receivers. These sensors are part of the International Monitoring System which is in place for the verification of the Comprehensive Nuclear-Test-Ban Treaty. They are positioned at a water depth of approximately 1 km, in the Sound Fixing and Ranging channel. This channel is a low-velocity layer that functions as an acoustic wave guide, thus facilitating very long range propagation with little attenuation. This study analyses transient signals between 2005 and 2018 triggered by submarine earthquakes and detected at station H10, situated near Ascension Island in the South Atlantic Ocean. This station consists of two three-element (triplet) arrays with an aperture of approximately 2 km each. The triplets are 126 km apart. Array processing techniques applied to individual triplets are prone to be biased by local conditions of the array surroundings. We demonstrate that this bias is largely suppressed when jointly processing both triplets as one six-elements array. Due to the malfunction of element S1, data quality decreased after October 2013 and our results are less robust. For the 2005-2013 period, we retrieve a temperature of 4.3 to 4.8 °C.