Subduction zones – major plate boundaries where one tectonic plate thrusts beneath another – host the largest earthquakes on Earth and thus pose major risks for earthquake damage and tsunami.  The Cascadia subduction zone (CSZ) has been the site of very large “megathrust” earthquakes in the past, created by the active subduction of the Juan de Fuca, Explorer and Gorda plates underneath North America. The potential for large megathrust earthquakes poses a major hazard to the population centers of the Pacific-Northwest United States. Currently, the plate boundary is mostly aseismic (devoid of small, regularly occurring earthquakes), making it difficult to map using common methods of earthquake location.  To minimize the risks to human lives and infrastructure, we need better knowledge of the geometry of the plate boundary, to improve models of seismic wave propagation in the region.  This means we need answers to such questions as, “How deep is the plate boundary offshore?”, “How slippery is the plate boundary – for example, is it lubricated by overpressured fluids?”, and “How does deformation transfer from the plate boundary into the overlying sediments off the coast?”

Our project focused on developing new methods to address those questions, by improving our ability to make subsurface images from marine seismic reflection data.  A major challenge in producing such subsurface images is “focusing” the image, in much the same way that the image in a microscope or telescope must be focused.  However, the problem is much more complex in reflection seismology – instead of having just one focusing know to turn (as in a telescope), in seismic imaging a better analogy would be thousands of independent focusing “knobs”, each of which could be turned independently.  Finding the right combination of “turns” on those knobs is a highly complicated problem, of the sort scientists call “nonlinear.”  We developed a new workflow for tackling this problem, by combining two sophisticated image processing methods that were previously unlinked:  waveform inversion and prestack depth migration.  We then applied this workflow to open-access seismic reflection data acquired off the coast of Washington state over the Cascadia subduction zone.

Our results showed new details of the geometry of the subduction system off the U.S. Pacific Northwest.  We imaged (1) a zone of low acoustic impedance underneath the Pleistocene accretionary prism; (2) a lack of a strong decollement reflection throughout the section; (3) discontinuous reflectivity from the subducting oceanic crust; and (4) a shallow dip (~1.5 – 4°) of the top of the subducting plate beneath the “accretionary prism” (sediments scraped off the incoming plate and squeezed by plate boundary forces).  We used our seismic velocities to create a model of how fluids are expelled from the prism and found that over the region covered by our seismic data ~750 km³ of fluid has been expelled from the accretionary prism at an average fluid expulsion rate of 1.1 mm/yr.

Finally, the new seismic processing workflow developed in this project should find wide use in the broader geophysical community, especially in areas of complex geological structures.  

Last Modified: 04/13/2016
Modified by: W. Steven Holbrook