A Systematic and Comprehensive Review on the Application of the Distributed Acoustic Sensing Data in Subsurface Application: Methodologies, Deep Learning Approaches, and Future Prospects

Distributed acoustic sensing (DAS) converts conventional fiber-optic cables into extensive seismic sensor arrays, providing unparalleled resolution for subsurface and seismic observations. This extensive study consolidates contemporary practices and methodological workflows in DAS-based subsurface monitoring, encompassing microseismic event detection, geothermal reservoirs, and carbon storage surveillance, with a focus on data-processing strategies and standardization initiatives. Eight main scientific databases provided 532 publications that were published between January 2014 and January 2025. After an extensive multistage screening process, 149 preprocessing, event detection, and classification experiments were selected for comprehensive investigation. DAS data processes were classified into principal topics, including denoising, normalization, segmentation, feature extraction, and machine learning. Convolutional neural networks (CNNs) have emerged as the predominant classification methods. Nonetheless, substantial deficiencies have been recognized in data annotation methodologies, benchmarking criteria, and incorporation of domain expertise in machine learning workflows. This analysis emphasizes the necessity for uniform processing pipelines and interpretable machine-learning models to progress the area. The results establish a robust basis for the advancement of next-generation data-driven subsurface sensing technologies utilizing DAS.