Enhancing 1D and 2D passive seismic imaging of urban ambient noise DAS recordings

Abstract

Ambient noise tomography Derived from Distributed Acoustic Sensing (DAS) deployed on existing telecommunication networks provides an opportunity to image the urban subsurface at local to regional scales and high resolution effectively with a small footprint. This capability can contribute to the assessment of the urban subsurface's potential for sustainable and safe utilization in countless applications, such as geothermal development of an area. However, extracting coherent seismic signals from the DAS ambient wavefield in urban environments remains a challenge. One obstacle is the presence of complex noise sources in urban environments, which may not be homogeneously distributed. Consequently, long-duration recordings are required to calculate high-quality virtual shot gathers, which entails significant time and computational cost.   In this study, we present the analysis of 15 days of passive DAS data recorded on a pre-existing fiber optic cable (dark fibers) running along an 11~km long major road in urban Berlin (Germany). We identify anthropogenic activities, mainly traffic noise from vehicles and trains, as the dominant seismic source and use it for ambient noise interferometry. To retrieve Virtual Shot Gathers (VSGs), we apply interferometric analysis based on the cross-correlation approach. Before stacking, we designed a selection scheme to carefully identify high-quality VSGs, which optimizes the resultant stacked VSG . Moreover, we modify the conventional ambient noise interferometry workflow by incorporating a coherence-based enhancement approach designed for wavefield data recorded with large-N arrays. We then conduct Multichannel Analysis of Surface Waves (MASW) to retrieve 1D shear-wave velocity models of the subsurface along consecutive portions of the array and validate them against local lithologic models. Finally, a 2D velocity model of the subsurface is obtained by concatenation of individual 1D velocity models from overlapping array subsections. The expansion into 2D requires an automatic identification of high-quality VSGs based on unsupervised learning, such as clustering, to exclude transient incoherent noise in the process of selective stacking.   The clustering results reveal distinct groups of VSGs that exhibit similar patterns. These distinct groups provide valuable insights into the temporal variations in human activities and allow a better understanding and interpretation of the recorded DAS ambient noise data. We find that recordings obtained predominantly during rush hour are viable for further processing and improve the accuracy of dispersion measurements, in particular for traffic-induced noise data. Moreover, the resulting 1D velocity models correspond well with available lithographic information. The modified workflow yields improved dispersion spectra, particularly in the low-frequency band (< 1 Hz) of the signal. This improvement leads to an increased investigation depth along with lower uncertainties in the inversion result. Additionally, these enhanced results were achieved using significantly less data than required using conventional processing schemes, thus opening the opportunity for reduced acquisition times and efforts.