Inversion of multimodal dispersion curves from distributed acoustic sensing measurements for subsurface imaging: A field case of Garner Valley, California

Abstract

Utilizing the surface-wave (Rayleigh-wave) dispersion properties to estimate subsurface S-wave velocity structure has become a popular method due to simple data collection and analysis procedure during the past decades. In particular, the promotion of distributed acoustic sensing (DAS) technique enhances its popularity due to the low-cost of ultra-dense observations. Moreover, under the same exploration requirement, the cost of the DAS data collection is much lower than that of the traditional acquisition implemented by the nodal seismographs. Therefore, it is necessary to expand the DAS application. Here, the multimodal surface-wave dispersion curves are successfully extracted from the phase-weighted stack of cross-correlation functions (CCFs) of the DAS data collected at Garner Valley Downhole Array field site, California by the proposed cylindrical-wave phase-shift. Then, the dispersion curves are inverted by the inversion workflow (expected to avoid the mode-misidentification) proposed by Yan et al. (2022a), consisting of staging strategy and pattern search with embedded Kuhn-Munkres (PSEKM) algorithm. Compared with other inversion schemes, the novel inversion method allows the presence of the observed values without prior mode-order definition in the inversion system, which omits the tedious and risky manual mode identification, and the mode-orders would be predicted dynamically during the inversion. The dispersion fittings for certain survey points indicate that the mode losses and aliasings appear to be widespread for the surface waves recorded on the free-surface. Finally, the S-wave velocity tomogram consistent with the geological materials of survey area is revealed. Our study demonstrates the feasibility of extracting the DAS-derived multimodal dispersion information and consequently obtaining a physical-reasonable subsurface S-wave velocity model.