Improving seismic interferometry for traffic noise by wavefield decomposition and deconvolution

Abstract

Traffic usually contains strong surface wave energy and can be acquired easily, and seismic interferometry algorithms are usually used to compute empirical Green’s function for surface waves to image subsurface shear wave velocity distribution. Due to the complexity of the traffic noise wavefield, estimation of robust empirical Green’s function usually requires hours or longer noise records. In this paper, we propose to improve interferometric extraction of surface wave Green’s function from traffic noise by deconvolution of the decomposed traffic noise wavefield, so that robust empirical Green’s function can be achieved with much shorter noise record acquired with linear arrays deployed along traffic roads. When the decomposed traffic noise wavefield contains unidirectionally traveling far-field surface wavefield, surface wave Green’s function can be estimated by a deconvolution. Crosstalk artifacts and near-field interference are excluded from the results, and virtual surface wave traces with a high signal-to-noise ratio can be computed with much shorter traffic noise traces. The proposed scheme is tested on a synthetic traffic noise dataset and a field traffic noise dataset acquired in a distributed acoustic sensing (DAS) experiment, and the results demonstrate that the proposed method generated higher signal-to-noise results than conventional interferometry schemes and that the empirical surface wave Green’s function computed from 4 seconds of traffic noise recording demonstrate comparable quality to active source data in the experiment.