EAGER: Using machine learning to develop a calibrated, remote sensing-based age model to improve late Quaternary slip-rate estimates in arid environments

本研究旨在改进地表地貌测年方法，进而改进确定断层滑动速率的方法。准确的滑移率对构造学和地震灾害研究至关重要，而且常常需要大量的地表年龄。由于缺乏可测定数据的材料、成本问题或现场的可及性，这种测定工作可能具有挑战性。最近的调查已将特定的遥感观测与地表地貌的绝对年龄直接联系起来。利用大型遥感数据库以及数据科学和机器学习的最新进展，本研究将整合多种不同类型的遥感数据和已发布的年龄数据，开发一个校准的年龄模型，该模型将应用于加州东南部的断层地貌。该方法将应用于东部Garlock断层，这是该地区的一个主要走滑断层，并有助于回答有关该断层在南加州构造中的作用的长期问题。这项研究将对加利福尼亚东南部许多活动断层的地震危险性分析做出重大贡献，该地区受到破坏性地震的威胁，人口超过300万。改进地震危险性评估对联邦、州和地方机构和管理机构、广泛的行业和公众都至关重要。本研究产生的模型也可以为未来世界范围内的地表年龄研究提供一个框架。此外，本研究将通过促进一名女研究生和至少两名本科生的教育和培训，以及两名早期职业研究人员（包括一名女助理教授）的专业发展，为STEM劳动力的发展做出贡献。地质滑动率是地震灾害分析的重要组成部分，对于解决构造学和地震学研究前沿的许多紧迫问题至关重要。然而，晚新生代和现在滑动速率的差异仍然存在争议，特别是当滑动速率估计跨越不同的活动时间尺度时。从观测偏差/限制中区分真正的滑移率差异需要对滑移率及其时空变异性有一个准确的（和自一致的）看法。然而，由于缺乏可追溯材料、成本问题或现场的可及性，获得稳定的滑动速率仍然具有挑战性。为了应对这些挑战，最近的调查将特定的遥感指数与地貌的绝对年龄直接关联起来。利用过去20年的广泛的遥感数据组合库以及数据科学和机器学习的最新进展，研究人员将在这些努力的基础上进行扩展，并将几种类型的遥感数据与已发布的地质年代学数据相结合，开发一个校准的地表属性-年龄模型，该模型将应用于加州东南部东加利福尼亚剪切带/南沃克Lane的断层地貌。整体模式将包含不同的模拟响应之间的遥感值和地表年龄。集成建模使用各种统计和计算模型来融合一系列单变量模型来解决分类和回归问题。有了各种各样的遥感器、波段和可用的空间尺度，丰富的数据可以整合到一个有凝聚力的、健壮的模型中。研究人员认为，与单个模型相比，这种严格的数据处理可能会大大改善结果年龄的不确定性。一旦实施，这项工作将产生一种校准方法，利用遥感数据和东部Garlock断层的新滑动率来估计地表年龄。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This study aims to improve the methods surrounding surface landform dating, and thus methods for determining rates of fault slip. Accurate slip rates are essential for tectonics and earthquake hazards research, and often require numerous surface ages. Such dating efforts can be challenging due to a lack of datable materials, cost concerns, or accessibility of field sites. Recent investigations have directly correlated specific remote sensing observations to the absolute age of surface landforms. Using a large repository of remote sensing data and recent advances in data science and machine learning, this study will integrate multiple, distinct types of remotely sensed data with published age data, to develop a calibrated age model that will be applied to faulted landforms in southeastern California. The methodology will be applied to the eastern Garlock fault, a major strike-slip fault in this region, and aid in answering longstanding questions about the role of the fault in southern California tectonics. This study will make a significant contribution to earthquake hazard analysis of many active faults in southeastern California, a region under threat of damaging earthquakes and with a population of more than 3 million people. Improved earthquake hazard assessments are critical for federal, state, and local agencies and regulatory bodies, a broad spectrum of industry, and the public. The model produced by this study can also form a framework for future surface age studies around the world. Additionally, this study will contribute to the development of the STEM workforce by advancing the education and training of a female graduate student and at least two undergraduate students, as well as the professional development of two early-career researchers, including a female assistant professor. Geologic slip rates are essential components of seismic hazard analysis and critical to addressing many pressing questions at the forefront of tectonics and seismological research. However, discrepancies of Late-Cenozoic and present-day slip rates continue to be debated, particularly when slip rate estimates span different timescales of activity. Discriminating true slip rate discrepancies from observational biases / limitations requires an accurate (and self-consistent) view of slip rates and their temporal and spatial variability. However, obtaining robust slip rates remains challenging, due to lack of dateable materials, cost concerns, or accessibility of field sites. Addressing these challenges, recent investigations have directly correlated specific remote sensing indices to the absolute age of landforms. Using the broad combined repository of remote sensing data from the past 20 years and recent advances in data science and machine learning, the investigators will expand on these efforts and integrate several types of remotely sensed data with published geochronology data to develop a calibrated surface property-age model that will be applied to faulted landforms in the Eastern California shear zone / southern Walker Lane of southeastern California. The ensemble model will incorporate different modeled responses between sensed values and surface age. Ensemble modeling uses a variety of statistical and computational models to fuse an array of single variate models to solve classification and regression problems. With the variety of remote sensors, bands, and spatial scales available, a wealth of data can be consolidated into a cohesive, robust model. The investigators believe such rigorous data treatment may yield significantly improved uncertainties on resulting ages compared with individual models. When implemented, the proposed effort will yield a calibrated means of estimating surface ages using remote sensing data and new slip rates for the eastern Garlock fault.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.