This paper utilizes prevalent deep learning techniques, such as Convolutional Neural Networks (CNNs) and Residual Neural Networks (ResNets), along with the well-established machine learning technique, Random Forest, to efficiently distinguish between common lithologies including coal, sandstone, limestone, and others. This approach is highly significant for resource extraction—such as coal, oil, natural gas, and groundwater—by streamlining the process and minimizing the need for the time-consuming manual interpretation of geophysical logging data. The natural gamma ray, density, and resistivity log data were collected from 22 wells in the mountainous region of Southern Sichuan, China, yielding approximately 70,000 samples for developing lithofacies prediction models. All the models achieved around 80% accuracy in classifying carbonaceous lithologies and up to 88% accuracy in predicting other lithologies. The trained models were applied to the logging data in the validation dataset, and the outputs were validated against geological core data, showing overall consistency, although variations in the classification results were observed across different wells. These findings suggest that deep learning techniques have the potential to develop a general model for effectively handling lithology classification with well logging data.

The 2008 M_w7.9 Wenchuan earthquake ruptured northwest‐dipping imbricate oblique reverse faults along the Longmenshan thrust belt at the eastern margin of the Tibetan Plateau, and developed one of the most complex coseismic rupture patterns for reverse faulting events ever reported in intraplate settings. To evaluate to what extent complex fault geometry and heterogeneities in initial stress can explain the characteristics of coseismic observations, we simulate spontaneous dynamic rupture propagation governed by slip‐weakening friction law on the geometrically complex multifault system loaded by heterogeneous stress. Our model reproduces many observed features including the multifault ruptures, locations of the maximum slip, Interferometric Synthetic Aperture Radar observations, source time functions and peak ground velocities. Our results suggest that the maximum horizontal principal stress in the northern part of Beichuan Fault (BFC) is rotated approximately counterclockwise compared to that in the southern part of BFC, yet this rotation is not sufficient to produce a supershear rupture. We infer that the fault core zone in the southwestern Beichuan Fault (BCF) may have been more severely damaged than the northeastern counterpart, which provides an explanation that the rupture in the northern BCF propagates faster than in the southern portion. Besides, we find that Coulomb failure stress changes on Wenchuan‐Maoxian Fault (WMF) may be counteracted by the aftershock slips of the Lixian Fault, which suggests Wenchuan earthquake could not significantly push the WMF closer to failure.

Several observations and simulations suggested that the supershear rupture induced by free surface invariably occurs in strike‐slip faults rather than in dip‐slip faults. To explore the mechanism of supershear transition induced by free surface on oblique faults, we simulate spontaneous dynamic rupture propagation governed by slip‐weakening friction on nonvertical planar faults. We find that oblique thrust faults are more likely to produce free‐surface‐induced supershear transition than oblique normal faults and significant asymmetry arises between the two opposite directions along fault strike. The supershear rupture may only occur within shallow depth or over the entire seismogenic depth, depending on geometric relations between the slip and rupture propagation vectors. Our results help explain the lack of universal observation of supershear earthquakes. Besides, the asymmetric ground shaking on the oblique faults allows us to predict the possible spatial distribution of the high seismic risk.

The transient electromagnetic method (TEM) is widely used for detecting low-resistivity areas ahead of tunnel. However, implementing the 2-D or 3-D inversion of whole-space geo-electric models is not feasible due to the narrow space within the underground tunnel and the nonuniqueness of TEM inversion. To solve this problem, we develop a fast inversion operator guided by deep learning (DL), which translates the time-domain TEM measurements directly into the spatial probability of water-filled anomalies’s position. Trained by synthetic data, our system shows impressive adaptability to predict water-filled anomalies when implementing different transmit currents, source waveforms, coil turn numbers, and abnormal body sizes. Compared to traditional 1-D tunnel TEM inversion, our system demonstrates less ambiguity, superior stability, applicability, noise resistance, and higher computational efficiency. The effectiveness of this method has been further confirmed by physical model experiments and field data. This inversion operator can support instantaneous TEM detection of low resistivity in the tunnel activities.

The 2008 Wenchuan Mw 7.9 mainshock caused catastrophic destruction to cities along the northwestern margin of the Sichuan Basin. This earthquake did not activate the Wenchuan–Maoxian Fault (WMF) on the hinterland side and the conjugate buried Lixian Fault (LXF), but they could experience large earthquakes in the future. We propose a systematic scheme to develop scenario earthquakes for active fault systems with insufficient constrain of 3D fault geometries. We first performed stress tensor inversion to constrain the regional stress field. Then, we developed a new method to constrain fault geometries by inverting long-term slip rates under the given regional stress and applied it to the WMF. We conducted a set of 3D dynamic earthquake rupture simulations on the WMF and LXF to assess the scenarios of earthquake rupture processes. Several fault nucleation points, friction coefficients, and initial stress states are assessed, the general rupture patterns for these earthquake scenarios are evaluated, and finally, we find the scenarios that could fall into three groups. Depending on initial conditions, the dynamic rupture may start in the LXF, leading to magnitude-7.0 earthquakes, or start in the WMF, then cascade through the LXF, leading to magnitude-7.5 earthquakes, or both start and arrest in the WMF, leading to around magnitude-6.5 or -7.0 earthquakes. We find that the rupture starting on the reverse oblique-slip jumps to the strike-slip fault, but the reverse process is impeded.Graphical Abstract