DASAttn DATASET

This dataset has been curated specifically to evaluate the performance of the DASAttn model, a self-supervised denoising framework tailored for Distributed Acoustic Sensing (DAS) seismic data. It includes a diverse collection of real-world seismic events recorded from two representative DAS deployments: the Ridgecrest area in California, USA, and the Xinfengjiang Reservoir in Guangdong, China. The dataset covers a broad spectrum of ambient noise conditions and sensor configurations, including land-based straight fiber layouts and underwater spiral deployments. Each sample is stored as a .npy file, representing a 2D wavefield with dimensions of channel count \u00d7 time steps. This high-resolution dataset supports robust testing of denoising algorithms across scenarios with varying signal-to-noise ratios, source distances, and geological environments. The dataset is intended for academic use in machine learning, geophysical signal processing, and DAS-based earthquake monitoring.