NOISE SUPPRESSION OF RECEIVER FUNCTIONS USING CURVELET TRANSFORM

Abstract

Suppressing the scattering induced by the laterally heterogeneous media is important for imaging the crustal structure and its anisotropy from Receiver Functions (RFs) based on the laterally stratified model. Although the scattering can be suppressed, to some degree, with stacking technique or low-pass filtering, these may lead to undesired waveform distortion, information loss or resolution reduction. To avoid these problems, we make use of the curvelet transform technique, which is developing rapidly in recent years, to reduce the scattering field in the RFs. Unlike exploration seismology, our major challenge comes from the spatially nonuniform sampling of RFs, caused by the spatially incomplete and uneven distribution of stations and events. To overcome these difficulties, we combine the compressed sensing theory with the curvelet-based denoising method to realize the denoising and wavefield reconstruction, simultaneously. To verify our idea, we have tested the denoising and wavefield reconstruction with synthetic RFs and then apply our method to the observed data at one of the IRIS GSN stations and the western Sichuan array, respectively. The results show that: 1) our method is efficient in suppressing the scattering induced by the lateral heterogeneity of the crust, which leads to great improvement of the signal-to-noise ratio and spatial traceability of the RFs. This is valuable for the waveform imaging of the crustal structure and anisotropic parameters from the RFs; 2) the missing data caused by the event distribution can be correctly reconstructed; 3) our method can be applied to either single station or seismic array observations, but it is more efficient in single station observation than the seismic array study.