ChannelSeg3D: channel simulation and deep learning for channel interpretation in 3D seismic images

Seismic channel interpretation involves detecting channel structures which often appear as meandering shapes in 3D seismic images. Many conventional methods are proposed for delineating channel structures using different seismic attributes. However, these methods are often sensitive to seismic discontinuities (e.g., noise and faults) that are not related to channels. We propose a convolutional neural network (CNN) method to improve the automatic channel interpretation. The key problem in applying the CNNs method into channel interpretation is the absence of the labeled field seismic images for training the CNNs. To solve this problem, we propose a workflow to automatically generate numerous synthetic training datasets with realistic channel structures. In this workflow, we first randomly simulate various meandering channel models based on geological numerical simulation. We further simulate structural deformation in the form of stratigraphic folding referred to as “folding structures” and combine them with the previously generated channel models to create reflectivity models and the corresponding channel labels. Convolved with a wavelet, the reflectivity models can be transformed into learnable synthetic seismic volumes. By training the designed CNN with synthetic seismic data, we obtain a CNN which learns the characterization of channel structures. Although trained on only synthetic seismic volumes, this CNN shows an outstanding performance on field seismic volumes. This indicates that the synthetic seismic images created in this workflow are realistic enough to train the CNN for channel interpretation in field seismic images.

Origin and significance of seismic discontinuities in the continental upper mantle: An interdisciplinary study

<p>              The mid-lithosphere discontinuity (MLD) and the lithosphere-asthenosphere-boundary (LAB) are two well-known seismic discontinuities in the continental upper mantle. Both MLD and LAB are present in most of the continents but at different depths and with different magnitude of velocity change and sharpness. Understanding the causes for these discontinuities including their regional variations is critical in inferring the evolution of the continents from geophysical observations on these discontinuities.</p><p>              Among various models, we focus on the elastically-accommodated grain-boundary sliding (EAGBS) model that provides plausible and unified explanations for the MLD and the LAB (Karato and Park, 2019). This model has a few testable predictions, and the main purpose of this talk is to review the current status of these tests.</p><ul><li>(i) One assumption of the EAGBS model is that EAGBS is enhanced by water. A recent paper by Cline et al. (2018) challenges this hypothesis by showing that water has no effects on attenuation in Ti-doped hydrated olivine. However, the relevance of the results on highly Ti-doped olivine to Ti-poor real upper mantle is unclear.</li> <li>(ii) A clear and unique prediction of the EAGBS is the presence of a peak in seismic attenuation at/near the MLD. However, inferring an attenuation peak in a narrow depth range is challenging and this hypothesis has not been tested.</li> <li>(iii) Another prediction of the “dry” version of the EAGBS model for the MLD is that although seismic wave velocity drops and there is a peak in attenuation, electrical conductivity does not change.</li> <li>(iv) If the MLD is caused by EAGBS, then materials below are in the “relaxed” state. This would explain the lack of large velocity drop at the LAB. However, the validity of this explanation depends on the pressure dependence of grain-boundary sliding. If pressure dependence of EAGBS is large, then the un-relaxed state will re-establish itself at a relatively shallow depth within the lithosphere. In this case, a deeper thermal transition to the relaxed state should produce stronger LAB than reported.  </li> </ul><p>We have conducted an interdisciplinary study to address these issues including mineral physics and seismology. We found that the addition of Ti modifies the defect-related properties of olivine and complicates the application of Cline et al. (2018) to actual upper-mantle conditions. We determined the pressure dependence of olivine grain-growth, from which we infer that the pressure dependence of grain-boundary sliding is small. Regarding the seismological test of attenuation peak, we forward-modeled surface-wave dispersion in a dispersive medium. Calculations show that the over-tones of Love waves are a key to detecting an attenuation peak near the GBS transition. Combined with a comparison of seismological studies (on velocity and attenuation) and MT estimates of electrical conductivity, we will have better constraints on the validity of the EAGBS model for the origin of the MLD.</p>

Seismic discontinuities in the lithospheric mantle at the Dead Sea Transform

SUMMARY The Dead Sea Transform (DST) was formed in the mid-Cenozoic, about 18 Myr ago, as a result of the breakaway of the Arabian plate from the African plate. Higher resolution information about the sub-Moho structure is still sparse in this region. Here, we study seismic discontinuities in the mantle lithosphere in the region of the DST using a modified version of the P- and S-receiver function methods. We use open data from permanent and temporary seismic stations. The results are displayed in a number of depth profiles through the study area. The Moho is observed on both sides of the transform at nearly 40 km depth by S-to-p and in P-to-s converted signals. The lithosphere–asthenosphere boundary (LAB) on the eastern side of the DST is observed near 180–200 km depth, which is according to our knowledge the first LAB observation at that depth in this region. This observation could lead to the conclusion that the thickness of the Arabian lithosphere east of the DST is likely cratonic. In addition, we observe in the entire area a negative velocity gradient at 60–80 km depth, which was previously interpreted as LAB.

Data-oriented constraint on the interpretation of S receiver function and its application to observations of seismic discontinuities in the lithosphere–asthenosphere system

SUMMARY This paper aims to improve the robustness of interpretation in the S receiver function (SRF), a technique commonly used to retrieve forward scattering of S-to-P converted waves (Sdp) originated from the lithosphere–asthenosphere system (LAS) beneath the stations. Although the SRF does not suffer interferences from backward scattering waves such as the first multiples from the Moho, one major drawback in the method is that Sdp phases can interfere with P coda waves and it is conceivable that these signal-generated noise may be misinterpreted as Sdp phase from the LAS beneath seismic stations. Through systematic analysis of full-waveform synthetics and SRFs from catalogued source parameters, we find that the strong P coda waves before the S wave in the longitudinal-component waveforms result in unwanted signal-generated noise before the S wave in the synthetic SRFs. If the mean amplitude of SRFs after the S wave is large, dubious signal-generated noise before the S arrival are strong as well. In this study, we honor the level of these unwanted signal-generated noise and devise data-oriented screening criteria to minimize the interference between P coda waves and genuine S-to-P converted waves. The first criterion is LQR, a direct measure of the amplitude ratio between longitudinal P coda waves and radial S wave in the waveform data. The second criterion is AMP, the amplitude of SRFs after the S arrival. We illustrate that these criteria effectively measure the energy level of mantle waves such as the SP wave. With synthetics and real data, we demonstrate the effectiveness of LQR and AMP criteria in minimizing these unwanted signal-generated noise in the stacked SRFs down to 1–2 per cent, improving detection threshold and interpretation of Sdp phases from seismic discontinuities in the LAS.