High-fidelity fault detection on seismic images is one of the most important and challenging topics in the field of automatic seismic interpretation. Conventional hand-picking-based and semi-human-intervened fault-detection approaches are being replaced by fully automatic methods thanks to the development of machine learning. We have developed a novel multiscale attention convolutional neural network (MACNN) to improve machine-learning-based automatic end-to-end fault detection on seismic images. The most important characteristics of our MACNN fault-detection method are that it uses a multiscale spatial-channel attention mechanism to merge and refine encoder feature maps of different spatial resolutions. The new architecture enables our MACNN to more effectively learn and exploit contextual information embedded in the encoder feature maps. We determine through several synthetic data and field data examples that our MACNN tends to produce higher resolution, higher fidelity fault maps from complex seismic images compared to those of the conventional fault-detection convolutional neural network, thus leading to improved geologic fidelity and interpretability of detected faults.

In this study, we explore whether the Parker and Baltay (2022) site response models for the Los Angeles (LA) basin region can improve ground‐motion forecasts in the U.S. Geological Survey ShakeAlert earthquake early warning system (hereafter ShakeAlert). We implement the peak ground acceleration and peak ground velocity site response models of Parker and Baltay (2022) in ShakeAlert via the earthquake information to ground‐motion (hereafter eqinfo2GM) module, which predicts ground motions from the estimated earthquake parameters of magnitude, rupture length, and location. The nonergodic site response models for the greater LA area were developed using ground motions from 414 M 3–7.3 earthquakes in southern California. We test nonergodic ground‐motion forecasts for five earthquakes in the LA area: the 1994 M 6.7 Northridge earthquake, the 2008 M 5.4 Chino Hills earthquake, the 2019 M 7.1 Ridgecrest earthquake, the 2020 M 4.5 South El Monte earthquake, and a synthetic M 7.8 earthquake on the southern San Andreas fault from the ShakeOut scenario, which was the basis of a statewide emergency response exercise. From the test results, we find that with the nonergodic site response applied, ShakeAlert not only alerts larger areas but can also result in longer warning times in LA region. In addition, the modified Mercalli intensity (MMI) ground‐motion predictions generated by the ShakeAlert eqinfo2GM module are improved in accuracy when compared with the corresponding ShakeMap ground‐truth MMI when the nonergodic site response model is applied.

With a detailed microscopic image of a rock sample, one can determine the corresponding 3-D grain geometry, forming a basis to calculate the elastic properties numerically. The issues which arise in such a calculation include those associated with image resolution, the registration of the digital numerical grid with the digital image, and grain anisotropy. Further, there is a need to validate the numerical calculation via experiment or theory. Because of the geometrical complexity of the rock, the best theoretical test employs the Hashin–Shtrikman result that, for an aggregate of two isotropic components with equal shear moduli, the bulk modulus is uniquely determined, independent of the micro-geometry. Similarly, for an aggregate of two isotropic components with a certain combination of elastic moduli defined herein, the Hashin–Shtrikman formulae give a unique result for the shear modulus, independent of the micro-geometry. For a porous, saturated rock, the solid incompressibility may be calculated via an “unjacketed” test, independent of the micro-geometry. Any numerical algorithm proposed for digital rock physics computation should be validated by successfully confirming these theoretical predictions. Using these tests, we validate a previously published staggered-grid finite difference damped time-stepping algorithm to calculate the static properties of digital rock models.