Signal Enhancement of Distributed Acoustic Sensing Data Using a Spectral Subtraction–Based Approach

Abstract Distributed acoustic sensing (DAS) is a fast-developing technology that has gained significant popularity in the seismological community, particularly for microseismic monitoring operations. Its unique capability to convert fiber-optic cables into dense seismic arrays offers numerous advantages over conventional seismic networks. This is particularly true for microseismic monitoring in extreme environments, such as deep geothermal wells, because of fiber’s resistance to high temperatures. However, seismic signals acquired with DAS systems are generally characterized by higher noise levels than those acquired with standard seismometers. When working with DAS systems, traditional frequency filtering methods often fail to recover weak signals, resulting in limited noise reduction performance. In this work, we introduce a robust denoising approach that adapts a spectral subtraction–based method used in speech signal processing to improve DAS data quality. We validated our denoising workflow using both realistic synthetic data and real microseismic events recorded at the Frontier Observatory for Research in Geothermal Energy Enhanced Geothermal Systems site (Utah, United States). The results obtained with both synthetic and real DAS data demonstrate significant improvements in signal enhancement, showcasing the effectiveness of our approach even under poor signal-to-noise ratio (SNR) conditions. This method outperforms traditional frequency filtering techniques, offering a promising solution to enhance DAS signals and enabling the detection of weak seismic events characterized by low SNR.