Illuminating the Transition Between Steady Sliding and Episodic Tremor and Slow Slip Using Low Frequency Earthquakes at the Downdip Edge

Using data from the Array of Arrays and CAFE experiments, we have identified eight Low-Frequency Earthquake (LFE) families on the subduction plate interface, under the Olympic Peninsula, Washington State. We analyze the time history of each during the time interval 2007-2012. The updip-most family (LFE1) only lights up during the well-known northern Cascadia Episodic Tremor and Slip (ETS) events that recur every 15 months. The recurrence intervals shorten from updip LFE1 to the downdip-most family (LFE4), which repeats every 14 days; 30 times more frequently. This presentation focuses on the downdip family. See the Sweet presentation, this session, for an analysis of the updip-most LFE family. LFEs from family 4 typically have durations of about one hour, with as many as 100 repeats during that time. Unlike their updip counterparts, they occur as discrete events without other LFEs or tremor visible during that time. They are strongly modulated by tidal shear stress. Twice as many LFEs occur during encouraging shear stress as during discouraging times. In contrast, these same LFEs occur when tidal normal stress is compressive which should inhibit slip. To reconcile LFE occurrence with favorable tidal Coulomb stress requires that the friction coefficient be less than 0.2 .This extreme sensitivity to very small shear stresses also suggests near lithostatic pore fluid pressures. We propose that the bursts of LFEs in this family correspond to discrete slow-slip events that occur with remarkable regularity. To add up to plate rates, each burst would correspond to a little more than 1 mm of slip, and each individual LFE to a little less than 0.1 mm, assuming all the slip occurs in the form of LFE activity and each LFE ruptures the same spot. One of these event sequences was captured by our 1-km aperture 80-element Big Skidder Array in 2008. Careful stacked correlation functions from 32 LFEs relative to a reference event showed S-P times varied only up to 0.02s, which translates to at most 200 m separation in source centroids projected onto the ray path. We take this to be a crude lower bound of the source size. In addition, by stacking the P-waveforms we find that the duration of the source time functions often exceeds 2 s. This suggests a rupture propagation velocity may be only 100 m/s, consistent with the notion that these are indeed very slow earthquakes. Combining the slip per LFE (.1mm) with a circular fault of radius 200 m, the moment magnitude of each event would be 1.9, and the stress drop would be only 50 kPa.