Suppression of Transients From Continuous Seismic Records Without Modifying the Amplitude of Ambient Noise

Ambient noise recordings are commonly used for seismic imaging but they are often corrupted by earthquakes and other unwanted transients. In order to extract useful information from ambient noise recordings, it is required to first clean the data. This article exploits multiresolution property of discrete wavelet transform (DWT) to localize transients in different subbands. In this article, ambient noise recordings are decomposed into multiple frequency bands (levels) using DWT. At each level, segments of unwanted transients present in the ambient noise recordings are identified using an energy estimator, and these segments are thresholded subsequently. Universal method of threshold value estimation is used to determine the threshold. Clean ambient noise data are retrieved by performing inverse DWT on wavelet coefficients that have been subjected to thresholding. The performance of the proposed method is tested using continuous ambient noise recordings in two different geographical regions: 1) USArray (U.S.) and 2) Himalaya-Tibet. Further, results obtained using the proposed method are compared with those obtained using conventional 1-bit and running absolute mean temporal normalization techniques. It is observed that the proposed method effectively removes the unwanted transients while preserves the ambient noise. Clean ambient noise data obtained using the proposed and the conventional methods, is processed further to estimate the empirical Green's functions (EGFs) which are comprised of information containing surface waves. It is shown that the arrival time of the surface waves obtained using proposed and conventional methods exactly matches, however, amplitude of the surface waves differ.