Testing Theoretical models for Earthquake Clustering using Cl-36 Cosmogenic Exposure Dating of Active Normal Faults in Central Italy

很难预测未来破坏性地震（>Ms 6.0）的时间，因为地震会及时发生在任何给定的断层群上（图1）。由于聚类导致的地震复发间隔的变化鲜为人知，因为我们缺乏关于任何给定断层在足够长的时间内发生多次地震的地震时间的数据。这种知识的缺乏阻碍了地震发生的概率预测，因为递归区间的内在变异性，即变异系数，CV，是输入概率计算的关键参数（例如Ellsworth et al. 1999）。我们所知道的是，可以通过测量宇宙36Cl浓度来确定单个断层上的大量古地震的年代，但这样的数据非常罕见。地震时，正断层的隆起会使埋在地表下的岩石暴露在宇宙射线中。由于宇宙粒子与岩石矿物中的原子相互作用，暴露在外的岩石随着时间的推移积累了36Cl。因此，36Cl的浓度揭示了地球表面岩石暴露的年龄。暴露年龄剖面的步骤揭示了过去抵消地面的大地震的时间。这种阶级化暴露年龄剖面提供的地震记录比历史地震记录长一个量级（例如，在意大利Velino断层上，一个36Cl剖面约束了12-18 ka以来的7次古地震[Palumbo et al. 2004；图2]）。自公元1349年以来的完整历史记录显示，自那以后只有1次地震（公元1904年，Ms 5.5）。Velino断层并不代表意大利其他活跃的正断层，因为它是一条滑动相对缓慢的新断层（自12-18 ka以来平均垂直运动约为0.3毫米/年），并且它正在成长，以连接两个自12-18 ka以来以1-2毫米/年滑动的长期存在的断层。我们的初步数值模拟表明，这种连接断层的CV值与主要的盆地边界正断层的CV值不同。我们建议收集现场36Cl数据并进行数值模拟，这将使我们能够测试如何以及为什么递归变异性由断层几何形状和断层生长/联系控制。意大利中部的断层是理想的，因为(1)适合36Cl测年的断裂带丰富（Palumbo et al. 2004, Papanikolaou et al. 2005），(2)断层几何形状和生长历史受到很好的约束（Cowie和Roberts 2001， Roberts和Michetti 2004）， (3) P.I.和co .与意大利政府地震机构建立了联系，并且可以使用爱丁堡和达勒姆的36Cl和断裂带测绘设备。(4) P.I.有现有的数值模型运行，已经显示了CV如何随断层几何形状和断层生长而变化；(5)我们有使用地震递归来绘制地震概率的经验（Roberts et al. 2004）。我们将收集新形成的连接断层和长期形成的活动断层的36Cl剖面，以建立与断层几何形状相关的CV空间变异性。我们已经确定了地表暴露仅通过地表断层而不是侵蚀/沉积循环的地点。激光测绘（LiDAR）和现场测绘将进一步限制样品位置并优化采样策略。数值模拟将限制为什么CV在新生长的连接断层和长期建立的活动断层之间变化的物理学。我们将迭代地比较场和数值结果，以约束在每种情况下决定递归的底层物理。我们将使用递归变异性CV计算地震概率，并绘制地震危险性及其在断层几何形状中的变化，利用我们的发现来评估全球地震危险性。

It is difficult to predict the timing of future destructive earthquakes (>Ms 6.0) because earthquakes on any given fault cluster in time (Figure 1). Variability in earthquake recurrence intervals due to clustering is poorly known because we lack data on the timing of earthquakes on any given fault over a sufficient time period that several earthquakes will have happened. This lack of knowledge hinders probabilistic forecasts of earthquake occurrence because the intrinsic variability in recurrence intervals, the Coefficient of Variation, CV, is a key parameter for input into probabilistic calculations (e.g. Ellsworth et al. 1999). What we do know is that large numbers of palaeo-earthquakes on individual faults can be dated by measuring the cosmogenic 36Cl concentration, but such data are very rare. Uplift along an active normal fault during an earthquake causes the rocks that were buried beneath the ground surface to be exposed to cosmic rays. The rocks that become exposed, accumulate 36Cl through time due to interaction of cosmic particles with the atoms within the rock minerals. Thus the concentration of 36Cl reveals the age of exposure of the rock at the Earth's surface. Steps in exposure age profiles reveal the timing of large earthquakes that offset the ground surface in the past. Such stepped exposure age profiles provide earthquake records that are an order of magnitude longer than historical earthquake records (e.g. a single 36Cl profile constrained 7 palaeoearthquakes since 12-18 ka on the Velino Fault, Italy [Palumbo et al. 2004; Fig. 2]). The historical record, which is complete since 1349 A.D., reveals only 1 earthquake since then (1904 A.D., Ms 5.5). The Velino fault does not typify other active normal faults in Italy, because it is a new fault that slips relatively slowly (~0.3 mm/yr average vertical motion since 12-18 ka), and it is growing to link 2 long-established faults that are slipping at 1-2 mm/yr since 12-18 ka. Our initial numerical modelling shows that such linking-faults have different values for CV than the main basin-bounding normal faults. We propose to collect field 36Cl data and conduct numerical modelling that will allow us to test how and why recurrence variability is controlled by fault geometry and fault growth/linkage. Faults in central Italy are ideal because (1) fault scarps amenable to 36Cl dating are abundant (Palumbo et al. 2004, Papanikolaou et al. 2005), (2) fault geometries and growth histories are well constrained (Cowie and Roberts 2001, Roberts and Michetti 2004), (3) the P.I. and Co.I's have established links with Italian Government earthquake agencies, and have access to both 36Cl and scarp mapping equipment in Edinburgh and Durham, (4) The P.I. has existing numerical model runs that already show how CV varies with fault geometry and fault growth, (5) we have experience in using earthquake recurrence for mapping earthquake probabilities (Roberts et al. 2004). We will collect 36Cl profiles from newly-growing, linking faults, and long-established active faults to establish spatial variability in CV related to fault geometry. We have identified sites where surface exposure has been solely through surface faulting and not erosion/sedimentation cycles. Laser-mapping (LiDAR) and field mapping will further constrain the sample sites and optimise the sampling strategy. Numerical modelling will constrain the physics of why CV varies between newly-growing, linking faults, and long-established active faults. We will iteratively compare field and numerical results, to constrain the underlying physics that determines the recurrence in each circumstance. We will calculate earthquake probabilities using the recurrence variability, CV, and map seismic hazard and how it varies across the fault geometry, using our findings to assess seismic hazard worldwide.