Assessing Active Fault Mapping Gaps in Southern California Using Co-Seismic Surface Rupture Characteristics

Active fault mapping is an essential tool for predicting future surface ruptures. However, many earthquakes occur along unknown or partially mapped faults, even in “well-mapped” seismically active regions. This phenomenon is particularly evident in Southern California, as demonstrated by several surprising events, including: Ridgecrest 2019, El Mayor Cucapah 2010, Hector Mine 1999, Landers 1992, and Kern County 1952. Following these earthquakes on unmapped faults, it is often possible to find evidence suggesting pre-earthquake ruptures with paleoseismological studies. Thus, gaps in fault mapping may result from a lack of visible surface ruptures or from subtle signs that are challenging to identify. Recognizing these faults, despite weak signals in the landscape, is crucial for better predicting future shallow earthquakes and their potential impacts on human infrastructure. To understand why evidence of surface ruptures may disappear in certain fault sections, it is essential to learn how these ruptures develop following an earthquake.

The advent of very high-resolution satellite imaging, combined with image correlation techniques, presents new opportunities for characterizing the morphology of co-seismic surface ruptures. This study aims to investigate whether a systematic relationship exists between pre-earthquake fault mapping and the characteristics of observed co-seismic surface ruptures. Specifically, we search to determine whether faults mapped before a rupture exhibit statistically different co-seismic displacements or near-field deformation characteristics compared to unmapped faults, and whether ruptures lacking clear pre-event geomorphological expression display distinct signatures. We begin by analyzing the co-seismic surface rupture of the 2019 Ridgecrest earthquake and comparing the rupture characteristics with pre-event fault mapping obtained from the USGS database. This analysis will then be extended to the 1992 Landers and 1999 Hector Mine earthquakes to evaluate the robustness and generality of the observed patterns across multiple large strike-slip events. For each earthquake, we construct dense datasets sampled along the surface ruptures, integrating morphological information derived from 2-meter resolution digital elevation models (DEMs) and displacement measurements obtained through 1-meter image correlation. We employ an unsupervised machine learning approach, specifically a hierarchical clustering, to group rupture segments based on their similarities across various parameters.

This methodology enables us to identify distinct classes of surface rupture behavior and evaluate how their distribution relates to pre-existing faults across different earthquakes. Our analyses reveal a strong correlation between the presence of pre-seismic geomorphic signal and lithology, as well as the intensity of co-seismic displacement. We found that more erosion-prone sediments and regions with smaller co-seismic displacement tend to show limited geomorphic expression prior to the earthquake. Additionally, some subtle pre-earthquake geomorphic signals can indeed be detected and mapped using very high-resolution satellite imagery. One initial approach to enhance fault mapping practices would be to utilize very high-resolution imagery, particularly in arid and sedimentary regions.