Novel Methods for Analyzing Problematic Geodetic Data: Applications to Coseismic Slip Modeling and Surface Deformation Mapping

Abstract

In chapter 2 we describe a novel estimation methodology for estimating coseismic slip models from geodetic observations of coseismic surface deformation. We apply our method to observations from the August 9, 2020, M$_{w}$ 5.1 Sparta, North Carolina earthquake captured by ALOS-2 and Sentinel 1A/B interferograms, along with differential LiDAR. Due to the relatively moderate surface deformation and the inherent uncertainties in the interferometric data, the interferograms displayed low signal-to-noise ratios. This issue was exacerbated by resolution discrepancies between the interferograms and the DEM used in processing, leading to the introduction of spurious signals related to local structures. Despite these challenges, coseismic slip was effectively modeled using our proposed method. Five coseismic slip models were developed: one based solely on the Sentinel 1A/B interferogram, another using both interferograms, a third that included all InSAR data while masking spurious signals from industrial areas, one based only on LiDAR displacements, and a final model that combined masked InSAR and LiDAR data. Our method produced a coseismic slip model for the Sparta earthquake that better aligns with existing ancillary information. This application highlighted the methodology's effectiveness in managing noisy and incomplete data, demonstrating its robustness and reliability in generating accurate models under challenging conditions.

Chapter 3 applies the same estimation strategy to coseismic surface deformation from the 2010 M 6.3 Jiashian and the 2016 M 6.4 MeiNong earthquakes in southern Taiwan recorded by GNSS and InSAR. Notably, neither earthquake produced surface rupture, leading to uncertainty about the associated nodal planes from seismological focal mechanisms. For the Jiashian event, the new coseismic model aligns well with existing focal mechanisms and models. However, for the MeiNong event, geodetic data suggested a shallower and more northerly slip surface than previous models indicated. This discrepancy highlights potential biases in earlier models due to constraints on fault geometry, particularly depth. The similarity in fault strikes and the proximity of inferred slip regions suggest both earthquakes likely occurred on a single fault structure that dips eastward and steepens with depth. This application underscored our methodology's effectiveness in producing more accurate models by integrating the estimation of distributed slip with the search for fault geometry. This holistic approach enhances the reliability of the results, allowing for better representation of the underlying fault structure.

In chapter 4 we outline a novel method for fusing geodectic data and demonstrate it's effectiveness on observations of coseismic surface deformation. Our fusion approach utilizes a semi-parametric latent factor model framework, which combines a parametric mapping of relationships between various geodetic observational modalities with a non-parametric representation of both observations and the latent deformation field. Our method provides a single, three component deformation field determined through multi-output Gaussian processes regression. This technique leverages the complementary features of different data types to filter out spurious signals, leading to a more accurate estimate of ground deformation. We applied this method to analyze coseismic deformation from the 2020 Sparta, NC, USA, and the 2016 MeiNong, Taiwan, earthquakes, fusing InSAR data with LiDAR for Sparta and GNSS data for MeiNong. The results demonstrated significant improvements in the precision and coherence of the deformation field.