Investigating the Seismicity of Eastern Marmara Using Machine Learning Algorithms

In this study we explore the seismicity along the Eastern Marmara Sea using machine learning techniques to improve the detection threshold and improve the earthquake locations. The Sea of Marmara comprises network of faults including the northern strand of the North Anatolian Fault, the Main Marmara Fault (MMF). MMF features a ~150 km seismic gap that did not rupture in the last 250 years. In addition to the MMF, other normal and strike-slip faults generate seismicity in the vicinity especially to the south of the Princes’ Islands.  It is therefore crucial to understand whether this seismicity is related to the MMF or other faults.

Here by employing a convolutional neural network detection and phase picking algorithm (Mousavi et al., 2020) and using a phase associator based on a grid search of locations (Zhang et al., 2019) we show that we can increase the detected number of earthquakes significantly and obtain a catalog with very low travel time residuals. Our primary objective is to acquire an improved earthquake catalog to facilitate subsequent clustering and focal mechanism analysis, thereby illuminating the underlying fault system responsible for these seismic events.

(This study is funded by TÜBİTAK Project No 121Y407).