Gabriel Fabien-Ouellet

Subsea permafrost is widespread across the polar continental shelf, and mapping its distribution is crucial for assessing the possible impact of its degradation in response to global warming. While seismic techniques have been useful in determining the lateral extent of subsea permafrost in the Canadian Beaufort Sea, are limited when it comes to estimating its vertical extent. However, recent advances in deep learning inversion have enabled the generation of more reliable large‐scale, depth‐resolved velocity and attenuation models. In this study, we employ a multi‐input, multi‐output neural network (NN) to evaluate the distribution of subsea permafrost at a regional scale in the Canadian Beaufort Sea. Application of our method resulted in the velocity and attenuation estimation of 15 seismic lines collected on the outer continental shelf. Our results reveal a gradual reduction in ice content and bonding effects toward the shelf edge. Our models show for the first time the high spatial variability of subsea permafrost near the shelf edge, bringing new insights and constraints into the dynamics of ice formation. Our findings are confirmed by well logs tied to seismic inversion, and support previous studies on subsea permafrost distribution in the region. Our method resolves the thickness of the upper permafrost layer with greater accuracy than previous studies and provides a practical framework for regional‐scale detection of various permafrost conditions. Finally, our work opens the way to a larger‐scale automated reprocessing of existing Arctic multichannel seismic data that could help refine the distribution of subsea permafrost across the continental shelf.

Submarine permafrost in the Canadian Beaufort Sea is relict terrestrial permafrost, which is continuously degrading since the change of thermal conditions induced by a marine transgression that followed the last glaciation. Permafrost degradation has a crucial socio‐ecological significance because its thawing can result in geohazards like landslides or an increase in greenhouse gas emissions. These consequences are mostly regulated by the state of ice in permafrost. In this study, we use marine multichannel seismic data to apply a diving wave tomographic inversion on the outer 50 km of the Canadian Beaufort Shelf. Due to the close relationship between seismic velocity and ice content, we are able to infer detailed information about the present submarine permafrost condition. We find a clear variability of permafrost occurrences between the inner and outer Canadian Beaufort Shelf. At the inner shelf, discontinuous ice‐bonding permafrost occurs extensively close to the seafloor but is interrupted by taliks. Within the outer ∼27 km of the shelf, ice‐bonding permafrost is absent in the upper sediments and its top has plunged to >200 m below sea level. These findings add new details to the current state of the degrading permafrost. In addition, we observe seismic anisotropy in the frozen permafrost sediments.

The dynamics of moisture content in cover systems constructed on mining wastes were monitored at the pilot scale using 2D autonomous, remote, and noninvasive time-lapse electrical resistivity tomography combined with conventional point sensors. A methodology was proposed to process the daily hydrogeophysical datasets from 23 m long instrumented sections of covers with capillary barrier effects (CCBEs) designed to act as oxygen barriers, and covers with low saturated hydraulic conductivity layers (LSHCCs) designed to limit the water infiltration rate. Hydrogeophysical monitoring suggested that CCBEs were able to maintain high degrees of saturation in the moisture-retaining layer throughout the 1 year monitoring period, which would make it an efficient oxygen barrier. Larger spatio-temporal changes in moisture content were observed in LSHCCs and most of the low hydraulic conductivity layers remained below 85% saturation, which was attributed to the combined effect of low precipitation, rapid vegetation development, and water percolation through the cover. The methodology proposed in this pilot-scale ”proof-of-concept” study allowed the hydrogeological behavior of the cover systems to be monitored in the 23 m long instrumented sections using continuous geoelectrical data, which demonstrated that this innovative monitoring technique could be useful for geochemical and geotechnical monitoring programs in large-scale mining waste storage facilities.

First-break picking is an essential step in seismic data processing. First arrivals should be picked by an expert. This is a time-consuming procedure and subjective to a certain degree, leading to different results for different operators. In this study, we used a U-Net architecture with residual blocks to perform automatic first-break picking based on deep learning. Focusing on the effects of weight initialization on this process, we conduct this research by using the weights of a pretrained network that is used for object detection on the ImageNet dataset. The efficiency of the proposed method is tested on two real datasets. For both datasets, we pick manually the first breaks for less than 10% of the seismic shots. The pretrained network is fine-tuned on the picked shots and the rest of the shots are automatically picked by the neural network. It is shown that this strategy allows to reduce the size of the training set, requiring fine tuning with only a few picked shots per survey. Using random weights and more training epochs can lead to a lower training loss, but such a strategy leads to overfitting as the test error is higher than the one of the pretrained network. We also assess the possibility of using a general dataset by training a network with data from three different projects that are acquired with different equipment and at different locations. This study shows that if the general dataset is created carefully it can lead to more accurate first-break picking, otherwise the general dataset can decrease the accuracy. Focusing on near-surface geophysics, we perform traveltime tomography and compare the inverted velocity models based on different first-break picking methodologies. The results of the inversion show that the first breaks obtained by the pretrained network lead to a velocity model that is closer to the one obtained from the inversion of expert-picked first breaks.

Marine seismic surveys can be used to map ice-bearing subsea permafrost on a large scale. However, current seismic processing technologies have limited capacity to image permafrost distribution at depth, mainly due to the low sensitivity of primary reflections and refractions to the velocity inversion found at the base of permafrost. Guided waves and multiples are more sensitive to the velocity variations below the top of permafrost, but they remain challenging to use in physics-based inversion approaches. A deep-learning-based seismic inversion has the potential to improve seismic imaging below the top of permafrost by automatically extracting information from all wave modes. We develop a multi-input neural network (NN) to estimate seismic velocities from marine seismic data. The network is trained on synthetic data generated from the representative distributions of the seismic properties of subsea permafrost. We find that our network can image large velocity contrasts and reversals in depth, typical of subsea permafrost. We use our network to estimate the P- and S-wave velocity and Q-factor models from a seismic line in the Beaufort Sea. The NN indicates highly discontinuous subsea permafrost with variable thickness in the area. Our work shows that deep-learning-based seismic inversion could become a cost-effective technology to map the distribution of subsea permafrost on a large scale and, more generally, high-velocity geologic layers located in shallow waters.

Estimating seismic attenuation from shallow geotechnical borehole surveys can be a delicate task. Measured data are often collected within the source near-field domain and the classic inverse-distance correction of the geometric spreading (GS) is inappropriate at this scale. We develop a novel approach based on a 3D full-waveform modeling to substitute the inverse-distance correction. It consists of scaling the picked amplitudes using their counterparts obtained from an elastic simulation carried out under conditions mimicking the data acquisition. The seismic attenuation may be inferred from the corrected amplitudes using either a piecewise regression or a ray-based inversion. Numerical experiments involving P- and S-wave synthetic data indicate that our correction better compensates for the GS effect than the inverse-distance correction. For a synthetic example with 5% noise, the Q-factor values derived from amplitude corrected via the proposed approach have a relative error of approximately 10% compared with 40% for the traditional correction. We investigate the effect of the velocity and density uncertainty upon the calculated correction terms and show that our approach is unbiased and stable. Finally, the robustness of our workflow is assessed on a real case study involving a P-wave data set acquired in a geotechnical borehole.

Monitoring the change in permafrost conditions and distribution is crucial for forecasting global warming. As seismic velocities increase with ice content, marine seismic surveys can map the top of ice-bearing subsea permafrost on a large scale. However, conventional seismic methods cannot fully map internal velocity variations linked to changes in ice content because the removal of guided waves and multiples is challenging for reflection processing. Nevertheless, these arrivals carry information about velocity variations with depth. We investigate if a joint analysis of various wave arrivals could provide information about velocity variations within permafrost and accurately estimate the permafrost thickness. Through a sensitivity analysis, we estimate the feasibility of using different wave arrivals, such as reflections, refractions, multiples, and guided waves, to better characterize permafrost conditions. We find that guided waves are sensitive to velocity variations within permafrost and can detect the depth of the permafrost’s base under certain geologic conditions. Then, we combine the analysis of all seismic arrivals to derive a subsurface permafrost model for a seismic line collected in the Beaufort Sea. The joint analysis reveals the transitions in depth between ice-bonded and partially ice-bonded permafrost and offers a crude estimate of permafrost thickness.

The Canadian Arctic Southern Beaufort Sea is characterized by prominent relict submarine permafrost and gas hydrate occurrences formed by subaerial exposure during extensive glaciations in Pliocene and Pleistocene. Submarine permafrost is still responding to the thermal change as a consequence of the marine transgression that followed the last glaciation. Submarine permafrost is still underexplored and is currently the focus of several research projects as its degradation releases greenhouse gases that contribute to climate change. In this study, seismic reflection indicators are used to investigate the presence of submarine permafrost and gas hydrates on the outer continental shelf where the base of permafrost is expected to cross‐cut geological layers. To address the challenges of marine seismic data collected in shallow water environments, we utilize a representative synthetic model to assess the data processing and the detection of submarine permafrost and gas hydrate by seismic data. The synthetic model allows us to minimize the misinterpretation of acquisition and processing artifacts. In the field data, we identify features along with characteristics arising from the top and base of submarine permafrost and the base of the gas hydrate stability zone. This work shows the distribution of the present submarine permafrost along the southern Canadian Beaufort Sea region and confirms its extension to the outer continental shelf. It supports the general shape suggested by previous works and previously published numerical models.