Geophysical Journal International

SummaryThis study focuses on the hydrophone component of a dense ocean bottom cable dataset from the North Sea. This data had already been used in the past to illustrate the high resolution power of full waveform inversion based strategies. We have developed a highly scalable implementation of a visco-acoustic full waveform inversion engine making it possible to double the frequency content of the inverted data compared to previous studies, using simultaneously up to almost 50,000 CPU. This results in a multi-parameter reconstruction of the subsurface, where the P-wave velocity, the density and the quality factor are reliably reconstructed down to 2 km depth, with a resolution of about 10 meters.

SummaryHarmaliére is an active landslide located in a mountainous region of southern France where the presence of a thick layer of clays provides favorable conditions for the development of slowly moving landslides. However, at Harmaliére, the alternation of sudden reactivation and quiet episodes suggests that specific structural and geomechanical properties control its kinematics. In order to shed light on its subsurface properties, we deployed for one month a dense network of seismic nodes within the landslide and recorded active-source and ambient noise seismic data. These datasets have been first independently processed with dedicated interferometry-based processing and inversion workflows to reconstruct P-wave (active sources) and S-wave (seismic noise) velocity models. Full wave-equation tomography is then performed to improve the reliability and resolution of the obtained elastic model by iteratively fitting the virtual gathers obtained by cross-correlation of the ambient-noise recordings. As opposed to conventional ambient-noise tomography, the approach fully accounts for topography and 3D elastic heterogeneities. The obtained high-resolution 3D models are then qualitatively interpreted in terms of landslide properties and geological lithologies, that can influence landslide kinematics.

SummaryDebris flows pose a significant threat to the sustainable development of mountainous regions. As an effective real-time sensing technique, microseismic monitoring plays a critical role in the detection and analysis of debris flow activity. However, current microseismic monitoring technologies face challenges in distinguishing mixed signals originating from different sources, limiting our understanding of the full dynamic evolution of debris flow events. To address this issue, we propose an unsupervised deep clustering-based signal classification framework, which focuses on analyzing the signal characteristics at various stages of debris flow events. A two-dimensional spectrogram dataset was constructed, encompassing signals from debris flows, rockfalls, earthquakes, and environmental noise. A deep autoencoder was employed to compress spectral features into a 16-dimensional latent space, followed by clustering using deep embedded clustering and Gaussian Mixture Models. Experimental results demonstrate that, after optimizing the feature space and data partitioning strategy, the proposed method achieves an average classification accuracy of 96.81 per cent across the four signal types. Power spectral density distribution analysis further confirms that this method not only accurately identifies debris flow signals but also effectively captures their energy distribution and dynamic evolution at different stages. Interpretability analysis reveals strong correlations between the extracted latent features and conventional seismological parameters, particularly the peak count of the time-domain autocorrelation function and the first quartile of the central frequency. Based on this method, a complete segmentation of debris flow events was successfully achieved, revealing the typical signal characteristics and temporal evolution of each stage. Cross-station validation indicates that the proposed framework demonstrates strong robustness and generalization across different monitoring locations. In addition, preliminary exploration of its integration with supervised learning suggests its potential applicability in real-time monitoring scenarios, offering a novel approach for debris flow early warning. This study presents an efficient and intelligent method for debris flow signal recognition and dynamic monitoring.

SummaryBayesian formulations of inverse problems are attractive due to their ability to incorporate prior knowledge, account for various sources of uncertainties, and update probabilistic models as new information becomes available. Markov chain Monte Carlo (MCMC) methods sample posterior probability density functions (pdfs) provided accurate representations of prior information and many evaluations of likelihood functions. Dimensionality-reduction techniques such as principal component analysis (PCA) can assist in defining the prior pdf and the input bases can be used to train surrogate models. Surrogate models offer efficient approximations of likelihood functions that can replace traditional and costly forward solvers in MCMC inversions. Many problem classes in geophysics involve intricate input/output relationships that conventional surrogate models, constructed using samples drawn from the prior pdf fail to capture, leading to biased inversion results and poor uncertainty quantification. Incorporating samples from regions of high posterior probability in the training may increase accuracy, but identifying these regions is challenging. In the context of full waveform inversion, we identify and explore high-probability posterior regions using a series of successively-trained surrogate models covering progressively expanding wave bandwidths. The initial surrogate model is used to invert low-frequency data only as the input/output relationship of high-frequency data are too complex to be described across the full prior pdf with a single surrogate model. After a first MCMC inversion, we retrain the surrogate model on samples from the resulting posterior pdf and repeat the process. By focusing on progressively narrower input domain regions, it is possible to progressively increase the frequency bandwidth of the data to be modeled while also decreasing model errors. Through this iterative scheme, we eventually obtain a surrogate model that is of high accuracy for model realizations exhibiting significant posterior probabilities across the full bandwidth of interest. This surrogate model is then used to perform an MCMC inversion yielding the final estimation of the posterior pdf. Numerical results from 2D synthetic crosshole Ground Penetrating Radar (GPR) examples demonstrate that our method outperforms ray-based approaches, as well as results obtained when only training the surrogate model using samples from the prior pdf. Our methodology reduces the overall computational cost by approximately two orders of magnitude compared to using a classical finite-difference time-domain forward scheme.