Components of Variability in Ground-Motion Model Derived from Distributed Acoustic Sensing

ABSTRACT Distributed acoustic sensing (DAS) allows meter-scale measurements along kilometer-scale fiber-optic cables, offering new opportunities to develop ground-motion models (GMMs) that resolve spatial variations in ground motion. However, developing GMMs from DAS measurements remains challenging because DAS observations are influenced by DAS-specific nonergodic components not present in seismometer-based measurements, including cable coupling and orientation. Moreover, channel correlation in DAS measurements can lead to biased uncertainty estimates of ground-motion variability when DAS channels are treated as independent stations; therefore, configuring correlation is essential for the reliable development of DAS-based GMMs. Here, we present the first regional, partially nonergodic DAS-based GMM using strain-rate data from a 400-channel array in Hualien City, Taiwan. We compile a dataset of peak strain rates and Fourier amplitudes (0.1–10 Hz) from 77 earthquakes (3.2≤Mw≤7.1, 45.75≤Rhypo≤170.80 km). Building upon classical seismometer-based GMMs, we extend the variability framework to consider (1) cable coupling associated with installation and environment types and (2) cable orientation. We further model (3) channel correlation inherent to DAS measurement principles and array geometry using Matérn kernels parameterized by both along-fiber channel and Euclidean distances. The resulting DAS-based GMM achieves magnitude–distance scaling comparable with classical GMMs, while allowing decomposition of ground-motion variability. We observe that the cable-coupling component exerts a broadband effect on ground motions and emerges as a dominant, frequency-dependent source of within-event variability. The orientation component captures frequency-dependent DAS directional sensitivity related to cable orientation and wavefront azimuth, and accounts for only a small portion of the variability. Introducing channel correlation substantially reduces channel-related standard deviations. Overall, our findings demonstrate that DAS-derived ground motions require a fundamentally different framework from classical seismometer-based GMMs, emphasizing the importance of deployment metadata and correlation modeling and providing a foundation for next-generation seismic hazard assessments using dense fiber-optic sensing.