Catalog-constrained models of tremor and slow slip

大约16年前，科学界开始意识到两种以前未被认识到的断层滑动类型。人们一直认为，深处的断层要么大部分时间都被锁住，只在短暂的地震中滑动，要么以板块构造的速度（每年几厘米）稳步移动。从2000年左右开始，人们认识到，沿着俯冲带断层的深部延伸，断层并不是简单地从浅层被锁定到在更深的地方以板块速度爬行。相反，在这种中间状态下，断层周期性地加速，滑动速度可能是板块构造速度的100倍，但随后又减速，不会产生地震。这些“间歇性慢滑事件”释放的能量与6.5级地震相当，但持续了几天到几周，而不是几秒钟。此外，这些缓慢的地震（用大地测量法观察到）伴随着地震仪上出现的一种新信号，称为“构造震动”。震颤由无数的“低频地震”组成，其能量释放相当于1 - 2.5级地震，但持续时间大约是地震的10倍（零点几秒）。间歇性慢滑和震颤的研究与地震危险有关，因为它们代表了俯冲带断层的浅层锁定部分能够产生9级地震的应力率增加的时间。也有人提出，地震的上倾程度可以用来评估未来大地震的下倾程度，这对西雅图的地面震动具有重要意义。但是，在我们能够自信地将这些想法转化为有关地震风险的陈述之前，我们需要更好地理解慢滑和震颤。提出的研究旨在提高我们对这些断层滑动类型的理解。当间歇性慢滑首次被发现时，问题是断层如何能够自发地、周期性地加速，但随后又减速而不产生地震。现在，我们的目标是利用对慢滑的观察来区分这种行为的不同假设。在过去的几年里，研究人员一直在改进地震检测/定位算法，目的是生成高分辨率的目录，扩大对慢滑动的观测约束（地震在这方面是有用的，因为它可以比潜在的慢滑动更准确地定位）。本建议的目标是建立以这些观测约束为指导的慢滑数值模型。主要观测结果包括：(1)虽然主震锋沿断层以5 ~ 10 km/d的速度横向传播，但在主震锋后面出现的次级震锋传播速度要快几十~几百倍。(2)这些次生偏移以快速地震反转的形式出现，它们都是从主锋面开始并向相反方向传播，或者沿主锋面传播而不受主锋面方向的影响。一些最大的快速地震反转开始于沿主要锋面的迁移。所有的迁移都不是从主要锋面开始，而是在长期传播方向上传播。(3)在主锋附近，地震迁移较小，且时间尺度短，不可能由潮汐驱动。随着这些时间间隔从几分钟逐渐增加到几小时，最终最大的快速地震反转被12.4小时的潮汐所调节。这项工作试图利用速率和状态依赖的摩擦模型再现这些观察结果，包括相对均匀的模型和那些明确包括旨在模拟震颤源的凹凸不平的模型。重点将放在分析地理解为什么模拟会这样做，这样他们的结果就可以超越他们特定假设的狭窄范围。

About 16 years ago the scientific community became aware of two previously unrecognized styles of fault slip. It had been thought that faults at depth either spent most of their time locked, slipping only in brief earthquakes, or else crept along steadily at the plate tectonic rate (centimeters per year). Starting around 2000 it was recognized that along the deep extension of faults in subduction zones, the fault did not simply progress from being locked at shallow depth to creeping at the plate rate at greater depth. Instead, in this intermediate regime the fault periodically accelerated to slip speeds perhaps 100 times the plate tectonic rate, but then slowed down again without producing earthquakes. These "episodic slow slip events" were comparable in energy release to magnitude 6.5 earthquakes, but lasted days to weeks rather than several seconds. In addition, these slow earthquakes (observed geodetically) were accompanied by a new signal seen on seismometers, termed "tectonic tremor". Tremor consists of myriad "low frequency earthquakes" that correspond in energy release to magnitude 1?2.5 earthquakes, but that last roughly 10 times as long (a few tenths of a second). Studies of episodic slow slip and tremor are relevant to earthquake hazards because they represent times of increased stressing rate on the shallow, locked portion of subduction zone faults capable of producing magnitude 9 earthquakes. It has also been proposed that the up-dip extent of tremor might be used to assess the down-dip extent of future great earthquakes, with important implications for ground shaking in Seattle. But we need to understand both slow slip and tremor better before we can confidently translate such ideas into statements about seismic risk. The proposed study seeks to improve our understanding of these styles of fault slip. When episodic slow slip was first discovered, the question was how faults could spontaneously and periodically accelerate, but then decelerate without producing an earthquake. Now, the goal is to use observations of slow slip to distinguish between competing hypotheses for this behavior. For the past several years the researchers have been improving tremor detection/location algorithms, with the aim of generating high-resolution catalogs that expand the observational constraints on slow slip (tremor is useful in this regard because it can be located more accurately than the underlying slow slip). The goal of this proposal is to develop numerical models of slow slip that use these observational constraints as guides. Key observations include: (1) Although the main tremor front propagates laterally along the fault at speeds of 5-10 km/day, secondary tremor fronts arise behind the main front that propagate tens to hundreds of times faster. (2) These secondary migrations take the form of rapid tremor reversals, all of which start at the main front and propagate in the opposite direction, or migrations that propagate along the main front independent of the main front orientation. Some of the largest rapid tremor reversals begin as migrations along the main front. None of the migrations start at the main front and propagate in the long-term propagation direction. (3) Closest to the main front, the tremor migrations are small and recur on timescales far too short to be tidally driven. As these intervals gradually increase from minutes to hours, eventually the largest rapid tremor reversals become modulated by the 12.4-hour tides. This work seeks to reproduce these observations using models of rate- and state-dependent friction, including both relatively homogeneous models and those that explicitly include asperities intended to mimic tremor sources. Emphasis will be placed on understanding analytically why the simulations do what they do, so that their results can be extrapolated beyond the narrow confines of their specific assumptions.