Ambient Vehicular Noise recorded on a 2D Distributed Fiber Optic Sensing Array :Applications to Permafrost Thaw Detection and Imaging

Distributed Acoustic Sensing (DAS) is a recently developed technique that allows the spatially dense ( 1m) continuous recording of seismic signals on long strands of commercial fiber optic cables. The availability of continuous recording on dense arrays offers unique possibilities for long-term timelapse monitoring of environmental processes in arctic environments. In the absence of a repeatable semi-permanent seismic source, the use of ambient surface wave noise from infrastructure use (e.g. moving vehicles) for seismic imaging allows tomographic monitoring of evolving subsurface systems. Challenges in such scenarios include (1) the processing requirements for dense (1000+ channel) arrays recording weeks to months of seismic data, (2) appropriate methods to retrieve empirical noise correlation functions (NCFs) in environments with non-optimal array geometries and both coherent as well as incoherent noise, and (3) semi-automated approaches to invert timelapse NCFs for near-surface soil properties.We present an exploratory study of data from a sparse 2D DAS array acquisition on 4000 linear meters of trenched fiber deployed in 10 crossing profiles. The dataset, collected during July and August of 2016, covers a zone of permafrost undergoing a controlled thaw induced by an array of resistive heaters. The site, located near a heavily used road, has a high level of infrastructure noise but exhibits distance-dependent variation in both noise amplitude and spectrum. We apply seismic interferometry to retrieve the empirical NCF across array subsections, and use collocated geophone and broadband sensors to measure the NCF against the true impulse response function of the medium. We demonstrate that the combination of vehicle tracking and data windowing allows improved reconstruction of stable NCFs appropriate for dispersion analysis and inversion. We also show both spatial and temporal patterns of background noise at the site using 2D beamforming and spectral analysis. Our results suggest that valuable information can be extracted from ambient noise recorded with DAS, particularly in the context of monitoring transformations in cold region environments.