Optimal Network Geometry for PBO

由于难以在考虑已知误差源和源复杂性的情况下表述问题，大地测量网的系统设计在文献中没有受到太多关注。考虑到所提出的网络的规模，确定最优网络设计对于板块边界观测站（PBO）来说尤为重要。随着大地测量社区开始PBO的安装阶段，通过修改站点位置来优化网络设计存在机会之窗。通过优化所提出的网络几何结构，PBO数据的质量将得到改善，并有望提高构造和火山事件的分辨率。根据PBO倡议的科学目标，正在通过一系列数值模拟来评估PBO的拟议网络几何形状。已经模拟了三个特征事件：圣安德烈亚斯断层中心的一次瞬变地震，卡斯卡迪亚俯冲界面的一次缓慢地震，以及圣海伦斯火山的一个传播堤防。瞬态源模型的结果将PBO网络性能描述为事件强度和持续时间的函数，并提供了对预期分辨率水平的见解。本工作还研究了岩浆输送的管道模型和沿板块边界的稳态变形。对于每个模拟，合成的GPS数据被创建，包括正式的白噪声和随机漫步组件。利用扩展网络反演滤波器从给定台站分布的合成时间序列数据中推断出原始源参数。扩展网络反转滤波器是卡尔曼滤波器的一种实现，它非常适合于从噪声数据中提取时间相关信号，并且在分析大型时间序列数据集时效率很高。通过将合成输入模型与滤波器估计的推断源模型进行比较，评估源过程的分辨率。额外的分析评估了最佳仪器组合（即GPS和钻孔应变计）和解决给定源过程所需的最佳站点密度。从这项工作中得出的建议将直接与PBO经理和常设委员会沟通。

The systematic design of geodetic networks has not received much attention in the literature because of the difficulty in formulating the problem while incorporating known error sources and source complexity. Identifying an optimal network design becomes particularly important for the Plate Boundary Observatory (PBO) given the scale of the proposed network. As the geodetic community begins the installation phase of PBO, there exists a window of opportunity to optimize the network design through the modification of station placements. By optimizing the proposed network geometry, the quality of data from PBO will be improved and hopefully allow for the greater resolution of tectonic and volcanic events.The proposed network geometry for PBO is being evaluated through a series of numerical simulations based on the scientific objectives of the PBO initiative. Three characteristic events have already been simulated: an aseismic transient on the central San Andreas fault, a slow earthquake on the Cascadia subduction interface, and a propagating dike at Mount St. Helens. Results for transient source models describe the PBO network performance as a function of the magnitude and duration of an event and provide insight on the expected level of resolution. Additional scenarios investigated in this work include conduit models of magma transport and steady-state deformation along the plate boundary. For each simulation, synthetic GPS data are created that include formal white noise and a random walk component. The Extended Network Inversion Filter is used to infer the original source parameters from synthetic time series data given the proposed station distribution. The Extended Network Inversion Filter is an implementation of a Kalman filter that is well suited to extract a time-dependent signal from noisy data and is efficient at analyzing large time-series data sets. Resolution of the source process is evaluated by comparing the synthetic input model with the inferred source model estimated by the filter. Additional analysis evaluates the optimal instrument ensemble (i.e. GPS and borehole strainmeters) and the optimal density of stations required to resolve a given source process. Recommendations derived from this work will be communicated directly with the PBO manager and standing committee.