PREEVENTS Track 2: 3D Nonlinear Simulation of Large Earthquakes on the Southern San Andreas Fault

加州最新的官方地震破裂预测（UCERF3）预测，在未来30年内，南加州发生7.7级或更大地震的概率为22%，圣安德烈亚斯断层可能是一个诱发断层。大洛杉矶地区的人口（超过1300万）和基础设施极易受到此类事件的影响。先前对南圣安德烈亚斯断层大地震情景的地面运动预测的好处受到底层模拟技术和带宽的缺陷的限制。该项目将利用超级计算资源的最新进展和复杂的数值模拟代码，将模拟提高到可用于工程设计的水平。由此产生的地面运动模型将允许改进地震危害分析，并具有在较高频率下被认为重要的真实模型特征。该项目将为显著改进地震危险性评估迈出第一步，最初是针对南加州圣安德烈亚斯断层上的一次大型地震。然而，这些结果可以用于未来可能发生大地震的其他地区（例如，加利福尼亚北部，华盛顿西部，Wasatch Front，新马德里地震带）。这项研究有望使对未来极端事件的峰值地面运动的精确预测成为可能，包括观测稀疏的近断层区域。如果使用得当，这些结果可能会影响当前的危险地图和工程设计，并减轻未来大地震造成的生命和财产损失。对南部圣安德烈亚斯断层大地震情景的大规模计算工作表明，由于三维盆地效应，所产生的长周期（超过0.5秒）地面运动水平具有显著的可变性。然而，从这些研究中预测的长周期地震动在实际用途上的应用有限，这是由于底层地震动建模的简化以及计算上对频率内容的限制。该项目将为工程设计和地震危害分析带来更复杂的模拟。以前的大型南圣安德烈亚斯断层事件的地面运动模型的最大频率将增加到5hz，具有真实的近地表速度，并包括在更高频率上被认为重要的新模型特征，如小尺度源和介质非均质性、频率相关的粘弹性衰减和场地效应。目前的工程实践中，特定场地的危害评估仍然依赖于一维非线性（或等效线性）模拟来预测土壤的响应。该项目将开发模型和计算策略，将非线性土壤响应集成到3D模拟中，使用钻孔阵列数据验证模型，并将模型预测应用于大型圣安德烈亚斯断层场景。此外，研究将探讨表面波如何促进放大，以及这种放大如何受到非线性和近地表非弹性衰减的影响。高度可扩展的基于gpu的有限差分代码AWP和可用的超级计算资源为具有挑战性的计算方面的研究提供了基础。该项目将完成正在进行的不连续网格能力的研究，以实现模拟中较低的近地表速度和较高的频率，在非线性介质中精确地插入源，并细化3D非线性流变学。

The latest official earthquake rupture forecast for California (UCERF3) predicts a probability of 22% within the next 30 years for a magnitude 7.7 or larger earthquake in southern California, with the San Andreas fault as a likely causative fault. The population (exceeding 13 million) and infrastructure in greater Los Angeles are highly vulnerable to such an event. The benefits of previous ground motion predictions for large earthquake scenarios on the southern San Andreas fault are limited by deficiencies in the underlying simulation techniques as well as bandwidth. This project will use recent advances in available supercomputing resources and sophistication of numerical modeling codes to bring the simulations to a level that can be used for engineering design. The resulting ground motion models will allow improved seismic hazard analysis with realistic model features deemed to be important at the higher frequencies. The project will provide a first step toward significantly improved seismic hazard estimation, initially for a large event on the San Andreas fault in southern California. However, the results can be used in other areas where large earthquakes are possible in the future (e.g., northern California, western Washington, Wasatch Front, New Madrid Seismic Zone). The research is expected to enable refined predictions of peak ground motions for extreme events in the future, including the near-fault area where observations are sparse. If properly used, these results could affect current hazard maps and engineering design, and mitigate the loss of life and property in future large earthquakes.Large-scale computational efforts for large earthquake scenarios on the southern San Andreas fault have shown significant variability of the resulting long-period (longer than 0.5 s) ground motion levels due to 3D basin effects. However, the long-period ground motions predicted from these studies have limited use for practical purposes, due to simplifications in the underlying ground motion modeling as well as computationally-imposed constraints on the frequency content. This project will bring the simulations to a sophistication useable for engineering design and seismic hazard analysis. The maximum frequency of previous ground motion models for large southern San Andreas fault events will be increased to 5 Hz with realistic near-surface velocities and include new model features deemed to be important at the higher frequencies, such as small-scale source and media heterogeneity, frequency-dependent viscoelastic attenuation, and site effects. The current engineering practice of site-specific hazard assessment still relies on 1D nonlinear (or equivalent linear) simulations to predict the response of soils. The project will develop models and computational strategies that integrate nonlinear soil response into 3D simulations, validate them using borehole array data, and apply the models predictively to large San Andreas fault scenarios. In addition, the research will investigate how surface waves contribute to amplification, and how this amplification is affected by nonlinearity and near-surface anelastic attenuation. The highly scalable GPU-based finite-difference code AWP and available supercomputing resources provide the foundation for the challenging computational aspects of the research. The project will complete on-going work on a discontinuous mesh capability to allow for lower near-surface velocities and higher frequencies in the simulations, accurate source insertion in a nonlinear medium, and refinement of 3D nonlinear rheology.