Pinning fault zone strength using small earthquakes in the Mendocino triple junction region recorded by a dense OBS array

Abstract:

The Mendocino triple junction contains a diversity of fault types including plate boundary strike-slip and thrust faults as well as intraplate faults within the subducting oceanic mantle that are expected to operate under significantly different rheological conditions. We analyze the data from the Cascadia Initiative’s Year-2 focused OBS array deployed at the triple junction region. We detect over 1000 earthquakes with magnitude ranging from 1 to 4.5 from Sep 2012 to Feb 2013. Locations refined with waveform cross-correlation arrival time measurements delineate two sub-parallel faults trending NW-SE revealing the complex geometry of the triple junction, as well as a intraplate fault possibly related to the 2010 M6.5 earthquake in the subducted Gorda plate. We are performing a joint inversion for 3D structure and hypocenter locations to further refine the image of the triple junction fault systems.

Our primary focus is on examining earthquake rupture mechanics in this complex fault network. Strength envelope calculations predict that the faults within the subducting Gorda plate support differential stress levels that are 1-2 orders of magnitude larger than what is typically assumed for the thrust interface. To determine if this contrast is reflected in the data, we estimate apparent stress for the M>2 earthquakes on the three types of faults. We compute displacement spectra using 2.56s time window from the picked arrival. Spectra with signal-to-noise ratio > 3 between 4 to 10 Hz are saved for further analysis. For the saved spectra, we apply two approaches using EGF method: (1) single event-pair deconvolution, where we select best-similar event pairs using relative locations derived from waveform cross-correlation, (2) iterative stacking deconvolution, where we solve for a event term, a station term and a path term using all the event-station pairs. Once the target event source spectrum is obtained, we compute the stress drop and apparent stress using the generalized “Boatwright” source model. The event durations will be independently measured from time-domain deconvolution using best-similar event pairs. We observe generally lower stress drops for earthquakes near the plate boundary and higher stress drops inside the Gorda plate, which is consistent with the expectation from fault strength profiles.