Comparison of Tectonic Tremor in Southwest Japan and Cascadia

Abstract:

Tectonic tremor in southwest Japan and Cascadia has a number of similarities and differences. We analyze tremor catalogs from these two subduction zones in order to highlight some of the differences and to determine whether the differences are related to data quality and methodologies. We start with existing tremor catalogs produced by Obara in Japan and Wech in Cascadia. The raw catalogs are constructed with a number of important differences related to data quality, time window size, horizontal versus vertical component data and use of amplitudes. The Japan catalog has finer temporal resolution and detects lower amplitude tremor. We have developed an algorithm to cluster the individual tremors into spatial and temporal swarms, which we interpret to represent slow slip events. We find that for both regions, the number of tremor swarms exceeding a given duration follows a power-law distribution. If duration, measured by number of fixed-length time windows of detected tremor during a swarm, is proportional to seismic moment, as has been commonly assumed, the moment inferred from these swarms follows a standard Gutenberg-Richter logarithmic frequency-magnitude relation with b-values close to 1. We find that for each catalog there is a clear region where this distribution explains the data well. In each case, as with all earthquake catalogs, there is a level below which the catalogs are incomplete. This level is lower in Japan than Cascadia. However, there is also a clear size above which the catalogs have significantly fewer events than expected by this distribution. This cutoff occurs at durations of about 5-25 hours for different geographic segments of southwest Japan and about 100 or more hours for Cascadia segments. We interpret this as representing a cutoff between unbounded, small events that follow the scale-invariant Gutenberg-Richter distribution, and large slip events that fill the down-dip width (or along-strike length) of the slow-slip zone and are therefore geometrically bounded. There is a strong correlation between this duration cutoff for a given segment and the down-dip width of the tremor zone, suggesting that local fault geometry controls the statistics of slow-slip events.