Distribution of Creeping Faults Based on a Global Search of Repeating Earthquakes

Repeating earthquakes occur on creeping faults due to stress concentration on locked seismic patches by fault creep in the surrounding area. Therefore, the repeating event sequences can be used to identify creeping faults and estimate their slip rate. There are many faults for which repeating earthquakes have been previously reported. However, the study areas are usually limited and the detection method varies from study to study. Here we investigate the global distribution of repeating earthquakes from a single detection method that is enabled by a previously published large systematic waveform cross-correlation dataset. The cross-correlation dataset by Dodge and Walter, 2015 (doi: 10.1785/0120140166) is from 310 million waveforms from a global seismic network for nearly 3.8 million events in 1970-2013. By using a magnitude-dependent frequency range and waveform similarity threshold to identify repeating earthquakes, we found more than 2500 repeating earthquakes, mostly above M 4.5 and belonging to about 600 sequences. The events are unevenly distributed along the worlds plate boundaries, suggesting variable interplate coupling. The NE Japan, Sumatra, Kuril, Ryukyu, Mexico, and San Andreas fault systems exhibit extensive repeating earthquake activity, thus supporting the existence of rapid fault creep. On the other hand, the Nankai, Cascadia, Alaska, and Mariana subduction zones had less repeating earthquake activity. These areas may have slower creep rate than the detection limit (1 3 cm/yr) or are devoid of repeating earthquake patches (i.e., they may be completely locked or effectively aseismic). We also found repeating earthquake activity in the subducted slab in the Hindu Kush subduction zone and along a few mid-ocean transform faults. These deep and far offshore examples capture active aseismic fault deformation that is otherwise difficult to measure by space geodetic methods. Acknowledgement: We thank Douglas Dodge for sharing the waveform cross-correlation data and his help during the analysis.