Developing high-resolution tremor catalogs to constrain numerical models of slow slip

Princeton University, Princeton NJ

Investigators

Abstract

One of the major discoveries in geophysics in the past decade has been that of ?episodic slow slip and tremor? in many of the world?s subduction zones. Most known faults either slip steadily at the plate tectonic rate (a few centimeters per year), or spend most of the time ?locked? and slip only during short-lived earthquakes, with slip speeds of order 1 m/sec and propagation speeds of order 3 km/sec (the sound speed of rock). Slow slip events, on the other hand, have average slip speeds of only ~0.1 micron/s and propagate up to 300 km along strike at remarkably reproducible speeds of roughly 10 km/day. Coincident in time and space with the geodetically-observed slow slip is a seismic signal termed ?tectonic tremor?. Unlike typical earthquakes, which have impulsive seismic wave arrivals, tremor is a low-amplitude signal that can last for hours and that most often lacks clearly identifiable seismic wave arrivals. Slow slip has now been discovered in nearly all subduction regions with sufficient instrumentation to see it, if it were present. Tremor is coincident in space and time with may slow slip events, and has also been discovered on the deep extension of the San Andreas fault in California. In addition to representing a previously unrecognized style of fault slip, episodic slow slip is relevant to seismic hazards because it increases the stressing rate on the locked portions of faults capable of producing magnitude 9 earthquakes. It has also been proposed that it may delimit the down-dip extent of slip during those earthquakes (bringing, notably, strong ground shaking considerably closer to downtown Seattle than had previously been thought). Because the resolution of geodetic data for deep faults is quite poor, tremor locations currently provide our most detailed images of the space-time history of slow slip. But because tremor lacks clear wave arrivals and can be active on multiple regions of the fault simultaneously, it cannot be located using standard techniques. We are developing a new tremor detection algorithm that, applied to the subduction zone off the coast of the Pacific Northwest, is currently producing the most accurate tremor locations in the world. Rather than the more traditional method of comparing seismograms from different time windows at the same station, it compares the same time windows at different stations. Relative location errors are often less than 1 km, allowing us to image in great detail secondary tremor fronts that arise behind, and propagate 1?2 orders of magnitude faster than, the main front of the slow slip event. By imaging these secondary fronts and their relation to the main with high fidelity, we expect to learn more about the processes underlying slow slip. In addition, studies of tremor are proliferating worldwide, but as yet there is no single, generally-agreed-upon method for locating it. By comparing our location method with more traditional methods and combining aspects of each, we stand to learn much about how to improve tremor location algorithms and how to apply this knowledge in other regions.