Noise suppression of microseismic data based on a fast singular value decomposition algorithm

Abstract

Raw data from microseismic monitoring is usually contaminated with ambient noise, which severely decreases the quality of the data and further affects the accuracy of arrival-based or amplitude-based source imaging performance., and thusThus, noise suppression of microseismic data plays an indispensable role in event detection during hydraulic fracturing. We investigatedeveloped a new way to effectively suppress noise from microseismic data via cross-correlation of traces from multiple microseismic channels. The cross-correlation between a reference trace and all available traces is computed to estimated the optimal time shifts between the waveforms in different channels. When the time shifts are estimated, the microseismic data can be rearranged to enhance the spatial correlation of the microseismic events. Next, a singular value decomposition (SVD) step is applied to extract the eigen-components or eigen-images of the noisy traces and to reject the noise in the data. When neighbor waveforms are too close, a windowed processing strategy is required to obtain well-flattened gather for the SVD processing. Considering that the SVD calculation is computationally expensive, we propose a fast SVD decomposition algorithm by derivation. The proposed method is easy to implement and can obtain robust performance in denoising multi-channel microseismic dataset. We validate the performance on both synthetic and real microseismic datasets. The proposed method can also be combined with active-source seismic data to efficiently suppress strong random background noise.