We simulate ground motion in southern California from an ensemble of 7 spontaneous rupture models of large (Mw7.8) northwest‐propagating earthquakes on the southern San Andreas fault (ShakeOut‐D). Compared to long‐period spectral accelerations from the Next Generation Attenuation (NGA) empirical relations, ShakeOut‐D predicts similar average rock‐site values (i.e., within roughly their epistemic uncertainty), but significantly larger values in Los Angeles and Ventura basins due to wave‐guide focusing effects. The ShakeOut‐D ground motion predictions differ from those of a kinematically parameterized, geometrically similar, scenario rupture: (1) the kinematic rock‐site predictions depart significantly from the common distance‐attenuation trend of the NGA and ShakeOut‐D results and (2) ShakeOut‐D predictions of long‐period spectral acceleration within the basins of the greater Los Angeles area are lower by factors of 2–3 than the corresponding kinematic predictions. We attribute these differences to a less coherent wavefield excited by the complex rupture paths of the ShakeOut‐D sources.
We describe an approach to model liquefaction extent that focuses on identifying broadly available geospatial variables (e.g., derived from digital elevation models) and earthquake-specific parameters (e.g., peak ground acceleration, PGA). A key step is database development: We focus on the 1995 Kobe and 2010–2011 Christchurch earthquakes because the presence/absence of liquefaction has been mapped so that the database is unbiased with respect to the areal extent of liquefaction. We derive two liquefaction models with explanatory variables that include PGA, shear-wave velocity, compound topographic index, and a newly defined normalized distance parameter (distance to coast divided by the sum of distance to coast and distance to the basin inland edge). To check the portability/reliability of these models, we apply them to the 2010 Haiti earthquake. We conclude that these models provide first-order approximations of the extent of liquefaction, appropriate for use in rapid response, loss estimation, and simulations.
Single image super-resolution (SISR) is of great importance as a low-level computer vision task. The fast development of Generative Adversarial Network (GAN) based deep learning architectures realises an efficient and effective SISR to boost the spatial resolution of natural images captured by digital cameras. However, the SISR for medical images is still a very challenging problem. This is due to (1) compared to natural images, in general, medical images have lower signal to noise ratios, (2) GAN based models pre-trained on natural images may synthesise unrealistic patterns in medical images which could affect the clinical interpretation and diagnosis, and (3) the vanilla GAN architecture may suffer from unstable training and collapse mode that can also affect the SISR results. In this paper, we propose a novel lesion focused SR (LFSR) method, which incorporates GAN to achieve perceptually realistic SISR results for brain tumour MRI images. More importantly, we test and make comparison using recently developed GAN variations, e.g., Wasserstein GAN (WGAN) and WGAN with Gradient Penalty (WGAN-GP), and propose a novel multi-scale GAN (MS-GAN), to achieve a more stabilised and efficient training and improved perceptual quality of the super-resolved results. Based on both quantitative evaluations and our designed mean opinion score, the proposed LFSR coupled with MS-GAN has performed better in terms of both perceptual quality and efficiency.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.