Slow-Slip Scaling Laws Inferred from Cascadia Tremor Swarms

Episodic tremor and slip (ETS) events, each with geodetically determined moment magnitudes in the mid-6 range, repeat about every 15 months under the Olympic Peninsula/southern Vancouver Island region. We have applied an automatic waveform envelope cross-correlation and clustering (WECC) algorithm to seven Cascadia-wide subarrays to search for non-volcanic tremor in 5-minute, 50% overlapping, time windows, revealing 70,000 tremor epicenters. The tremor epicenters cluster in time and space into nearly 200 tremor swarms. The number of hours of tremor per swarm ranges from about 1 to 470 hours. The smaller (inter-ETS) tremor swarms generally locate along the downdip side of the larger ETS swarms and occur much more frequently. In northern Washington, which is currently best monitored, the ETS events, as well as the larger inter-ETS tremor swarms initiate downdip and propagate updip. For the large ETS events, tremor swarm duration is proportional to geodetically determined seismic moment. We consider tremor swarms to be a proxy for slow slip for the smaller events as well, even though slip would be below current geodetic detection thresholds. An interpretation of the observed transition from longer duration, less frequent tremor swarms up dip to smaller more frequent tremor swarms down-dip, in terms of fault strength is the subject of a presentation by Wech. The combined inter-ETS and ETS swarms follow a power law relationship such that the number of swarms, N, exceeding duration τ is given by τ -0.66. If we assume that seismic moment is proportional to τ, as proposed by Ide et al. [Nature, 2007], we find that the tremor swarms follow a standard Gutenberg-Richter logarithmic frequency-magnitude relation, log10 N ≈ -bMw, with b = 1.0, which lies in the range for normal earthquake catalogs. Finally, crude estimates of the spatial dimensions of tremor swarms L suggest that L ≈ τ 1/n where n is between 2 and 3. A value of 2 is consistent with slip propagation rates being controlled by a diffusional process. In contrast, n is observed to be about 1 for normal earthquakes because rupture generally propagates at a velocity close to the shear-wave speed.