Slip heterogeneity of a subduction zone from detailed repeating earthquake analysis. A 15–year–long story from northern Chile

我提出了一个研究项目，系统地和定量地研究智利北部俯冲带的重复地震活动。RE序列是具有重叠破裂区域的事件群；它们通常通过几乎相同的波形来识别，这表明它们的震源机制和位置一致。REs的研究主要是由于它们在深度上约束运动（断层蠕变）的特殊潜力。这项工作的第一个科学目标将是沿板块界面的REs，计划对其进行分析，以阐明当前深度滑动速率的分布，并提供有关大逆冲地震旋回背景下锁定模式的宝贵信息。第二个主要目标将是位于俯冲板块深处的REs的机制。时至今日，它们的出现仍然是个谜。我计划研究时间和空间模式，并整合岩石物理约束来测试不同的假设机制（相互连接的孔隙网络的作用，脱水反应，剪切加热），从而有助于更好地理解中深度地震活动性。为了实现这两个目标，编写一个长期、完整和准确的RE目录是至关重要的。我将通过使用模板匹配和基于机器学习的工具来解决这个问题，从智利北部现有的高分辨率事件目录开始。为了尽量减少重复激活同一位置的地震关联的错误，我将改进RE识别的方法。其理念是结合波形相似性、差分走时测量的高精度相对事件位置，以及地震矩和应力降对震源尺寸的事件预测。在接下来的步骤中，计划使用最先进的和基于ml的提取、去噪、关联和定位程序来改进现有的地震目录。智利北部的拟议研究区域特别令人感兴趣，因为它被认为是一个地震间隙，有可能发生严重的大型逆冲事件。2014年，该地区经历了MW8.1级的伊基克特大逆冲地震，仅释放了部分储存的破裂潜力。自2006年以来，IPOC网络对该地区进行了调查，并编制了一份高分辨率地震活动目录，记录了超过18.2万次地震，包括中震和余震系列。在我最近的项目中，我对整个智利北部俯冲带的应力降变异性进行了研究，并为整个俯冲带创建了第一个连续的、全面的应力降图。这张图现在可以作为稀土分析的重要输入，以约束它们的破裂结构。请注意，我已经为本研究提出的基于机器学习的拾取和去噪、模板匹配和RE表征技术做了大量的准备工作。因此，我相信它的可行性。

I propose a research project to systematically and quantitatively study repeating earthquake (RE) activity in the northern Chilean subduction zone. RE sequences are groups of events with overlapping rupture area; they are usually identified by their nearly identical waveforms, which indicate coinciding source mechanism and location. REs are studied primarily due to their specific potential to constrain motion at depth (fault creep). The first scientific objective in this work will be REs along the plate interface, which is planned to be analyzed to illuminate the current distribution of slip rate at depth and to provide valuable information on the locking pattern in the context of the megathrust earthquake cycle. The second prime objective will be the mechanism(s) of REs that are located deeper within the subducting plate. Their occurrence as of today is still enigmatic. I plan to study the temporal and spatial patterns and integrate petrophysical constraints to test different hypothetical mechanisms (role of interconnected pore networks, dehydration reactions, shear-heating) and therewith contribute to a better understanding of intermediate depth seismicity, in general. For both aims, the compilation of a long-term, complete and accurate RE catalog is fundamental. I will approach this by using template matching and machine-learning based tools, starting with the existing high-resolution event catalog for northern Chile. To minimize errors in the association of earthquakes repeatedly activating the same location I will refine the methodology of RE identification. The idea is to combine waveform similarity, high-precision relative event locations from differential travel time measurements and event-wise estimates of source dimensions from seismic moments and stress drops. In a subsequent step it is planned to improve the already existing seismic catalog with state-of-the-art and ML-based picking, denoising, association and location procedures. The proposed study region in northern Chile is of particular interest as it is considered a seismic gap hosting the potential of a severe megathrust event. In 2014 the region experienced the MW8.1 Iquique megathrust earthquake which only released part of the stored rupture potential. The area has been surveyed by the IPOC network since 2006 and a high resolution seismicity catalog with over 182,000 earthquakes including for- and aftershock series. In my latest project, I have conducted a study on stress drop variability throughout the northern Chilean subduction zone and created the first continuous, comprehensive stress drop map for an entire subduction zone. This map can now serve as important input for RE analysis to constrain their rupture structures. Please note that I have already made substantial preparatory work for the machine learning based picking and denoising, template matching and RE characterization techniques proposed for this study. Therefore, I am convinced of its feasibility.