Qseek: A data-driven Framework for Automated Earthquake Detection, Localization and Characterization

We introduce a data-driven method and software for detecting and locating earthquakes in large seismic datasets. By combining seismic phase arrival annotations, delivered by neural network phase pickers, and waveform stacking with an adaptive octree search, we can automatically detect and locate seismic events even in noise-dominant seismic data. The resolution of the search volume is iteratively refined toward the seismic source location; this strategy facilitates an efficient, fast, and accurate search. We present a user-friendly and high-performance open-source software framework based on established frameworks, featuring event detection in layered 1D and complex 3D velocity models and event feature extraction capabilities, such as moment and local magnitude calculation from peak ground motions. We incorporated station-specific corrections and source-specific station terms into the search to enhance the location accuracy. We demonstrate and validate our approach by extracting extensive earthquake catalogs from large seismic datasets in different regions and geological settings: (1) Reykjanes Peninsula, Iceland; (2) Eifel volcanic region, Germany; and (3) Utah FORGE, USA. We capture seismic events from tectonic activity, volcanic swarms, and induced microseismic activity with magnitudes ranging from -1 to 5. Such precise and complete earthquake catalogs contribute to the interpretation and understanding of otherwise hidden subsurface processes.