Empirical subspace detection applied to triggered seismicity in the Sea of Marmara, Turkey
This code is part of the following paper
https://www.sciencedirect.com/science/article/pii/S0098300421000480
Abstract The northern branch of the North Anatolian Fault in the Sea of Marmara (SoM) shows a complex seismological behavior including shallow, fluid-related seismicity, deep seismicity, and locked and creeping segments. On July 25, 2011, a Mw 5 earthquakes occurred at a depth of 11.5 km in the western SoM and immediately followed by hundreds of triggered micro-earthquakes mostly outside the rupture area of the mainshock. We here present the application of empirical subspace detector to improve the detection of small events belonging to this sequence and compare these results with the match-filter technique. Using first the station with the highest signal-to-noise ratio, three main clusters are obtained with one cluster of 27 events and two clusters of 3 events leading to 1056 child-events detected. The child-events were manually selected and picked, and 457 child-events were located with NonLinLoc. The empirical subspace detector leads then to 5.5 times more events located compared with the number of parent events, and 1.6 times more events located than the match-filter technique. The subspace detector technique is sensitive mostly to the frequency band for filtering and the clustering parameters. Small changes in these parameters can lead to significantly different detection results. Extending the subspace detector to include network subspace statistics instead of a single station allows the inclusion of different stations and components, and should decrease the important false detection rate observed. The application of the empirical subspace detector to this sequence confirms the potential of this technique to improve event detection and better define spatio-temporal seismicity patterns.