Much of my recent work has involved studying slow slip and tectonic tremor in subduction zones.  Discovered about the year 2000 (in leap-frog fashion) in Japan and Cascadia, the two coupled styles of fault slip occur quasi-periodically, down-dip of the locked portion of subduction zone thrust faults and several strike-slip faults around the world.  Although when first discovered the question was “How can fault slip accelerate without leading to an earthquake?”, now that there are several proposed mechanisms that could plausibly generate episodic slow slip, that question has morphed into “How can we distinguish among the proposed physical mechanisms for slow slip?”.  I am tackling this question through a combination of numerical analysis and observations.

Jessica Hawthorne first used Earthscope borehole strainmeter data to show that the moment rate of slow slip in Cascadia is modulated by tidal stresses with amplitudes of only 1 kPa.  At the period of the strongest tide, 12.4 hours, the moment rate varies by about 25% above and below the mean, in phase with the tremor rate. The next step was to use this observation of modulation at the tens-of-percent level (as opposed to a few or nearly 100%) as a constraint on numerical models of slow slip.  Using what is arguably the simplest of the proposed mechanisms (a transition from velocity-weakening to velocity-strengthening behavior at a slip speed of about 1 micron/s), she came up with the first prediction of what controls the slow-slip recurrence interval for any of the proposed mechanisms.  She found that it was possible to match both the observed stress drops (or equivalently the recurrence interval) and tidal modulation, given sufficiently low effective normal stresses, but that to do so required pushing the limits of parameter space more than one might like.  

One lesson I learned from this work is that we need even more observations to judge between the proposed mechanisms for slow slip.  The most promising path, I think, lies in obtaining more accurate and complete tremor catalogs.  Tremor is notoriously difficult to locate because it lacks identifiable impulsive P-wave and S-wave arrivals, and is likely made up of simultaneous sources coming from multiple regions of the fault.  I am working on developing a “cross-station” detection/location algorithm, which compares the same short time window at different stations, as opposed to more traditional “cross-time” methods that compare different time windows at the same station.  Figure 1 shows how coherent the seismic signal can be at stations tens of kilometers apart.

This high degree of coherence has resulted in a tremor catalog that is more accurate than any other from anywhere in the world, with relative location errors in the 0.5­­–1 km range.  This has allowed us to image in unprecedented detail small-scale tremor migrations that piggy-back on top of the main slow slip event.  These tend to (a) start at or within about 1 km of the main tremor front, and propagate back along strike at rates 25-50 times faster, about 10-20 km/hr; (b) less commonly do the reverse, ending at the main front; or (c) propagate up- or down-dip at or within 1-2 kilometers of the main front.  Several examples of these secondary fronts can be seen in Figure 2 below, which shows a 10-km-wide region that was very active in each of the slow slip episodes in 2003, 2004, and 2005.  These images are for the 2005 event; the main front propagates SE to NW at about 10 km/day (this can be seen from the progression of the blue colors from panel to panel).

Activity as first the main front and then the secondary fronts pass through can best be seen on “space-time” plots such as in Figure 3, which shows both the slow progression of the main front to the NW and the much more rapid tremor “bursts” behind, for two days during each of the 2003 and 2004 slow slip episodes (in each of the 2003-2005 events the region of Figure 2was most active for about 2 days).  Colors in these plots indicate the relative “radiated energy” of the tremor detection.  At each location in each of the 3 episodes the tremor amplitude generally starts out low and progressively increases over a period of about ½ day before leveling off, spanning a range of nearly 3 orders of magnitude.