Abstract
Episodic tremor sequences comprised of overlapping low frequency earthquakes (LFEs) occur frequently along the San Andreas Fault. Accompanying slow slip activity has been detected from Global Navigation Satellite System (GNSS) data, confirming occurrence of the episodic tremor and slip (ETS) phenomenon here.
The characteristics of slow slip events (SSEs) impede comprehensive detection, making it challenging to study their occurrence patterns. We utilise extensive LFE data from a long running high resolution seismic network to gain insights into this more frequent and easily detectable aspect of the ETS process, with the aim to have a detailed understanding of the occurrence patterns and properties of LFEs. This will strengthen methods for the detection and modelling of SSEs.
Hidden Markov models were used to study the occurrence patterns of LFE events. Based on these models, LFE events along the San Andreas Fault can be classified into different states. Each state is a proxy for changes in the generating mechanisms that give rise to LFE events, with potential contributors including pore pressure and fault stress. We use the classification to illustrate a detailed picture of temporal changes in LFE activity - including the effects of events such as the 2004 Parkfield earthquake, and to highlight the diverse behaviours displayed across generating locations.
The evolution of LFE activity over space and time gives additional insights into how slow slip may propagate. We use clustering methods to reveal patterns in the migration of activity between spatially distinct generating locations, and identify locations with similar characteristics that are likely influenced by the same generating circumstances.