Implications and Limitations of Tremor as a Proxy for Slow Slip

Abstract

This thesis seeks to integrate geodetic and seismic observations to explore the relationship between tremor and slow slip on subduction zones. In particular, I evaluate the one-to-one relationship of tremor and slip in space and time, and test various hypotheses that describe their interaction and scaling. This work adds insight into the state of the Cascadia subduction zone and the slow-slip seismic cycle. In the first chapter, I use the surface displacements measured by GPS stations to analyze six major episodic tremor and slip (ETS) events from 2007 to 2016 in northern Cascadia and invert for slip on a realistic plate interface. Tremor is typically constrained to a relatively narrow band along dip that is downdip of the inferred locked megathrust. My results indicate that slow slip extends updip of tremor by about 15 km beneath the Olympic Peninsula. Additionally, I find that along-strike variations in the amount of slow slip updip of tremor correspond to changes in lithology of the overlying crust. In these ETS events, slow slip extends from the downdip portion of the tremorgenic region beyond the updip extent of tremor, although still downdip of the inferred locked megathrust. Slip updip of tremor is a persistent feature of all six ETS events at this along-strike location. Inversions that restrict slip to occur only in regions that generated tremor produced slip distributions with unphysical characteristics and unsustainable concentrations at the updip part of the tremor footprint. Updip slow slip without tremor may suggest that the gap between stress and strength widens updip above the observed limit of tremor. In these ETS events, the regions updip of tremor may undergo ductile failure surrounding potentially tremorgenic patches. A widening gap between stress and strength in the updip direction is consistent with an observed along-dip dependence of LFE occurrence and numerical simulations of slow slip. Alternatively, rheological properties in the region updip of tremor may favor stable slip and not permit seismic slip (i.e. tremor). In the second chapter, I explore the evolution of slow slip on the Cascadia megathrust during two large ETS events and compare stress changes to the spatial evolution of tremor from PNSN tremor locations. I use displacement time series from GPS stations, along with the Extended Network Inversion Filter to solve for the time-dependent fault slip on the megathrust. The 2010 (Mw 6.8) and 2012 (Mw 6.8) slow slip events propagated northward and southward, respectively, allowing us to assess directional effects on slip behavior. I observe that tremor occurs on the leading edge of propagating slipping regions, well ahead of the highest slip rates, independent of the along-strike propagation direction. Using the tremor distribution to generate synthetic surface displacement data, resolution tests show that the result of peak tremor rates leading peak slip rates is not due to biases introduced by temporal smoothing. Calculated stress changes due to the time-dependent fault slip distributions imply that tremor is sensitive to kPa of stress, consistent with studies of tidally-triggered tremor. Within the resolution of our model, these results are consistent with the hypothesis that significant tremor is triggered by stresses ahead of the highest slip rates. I also observe ongoing slip continuing several days after tremor has passed. Our observations are consistent with some numerical models of tremor patches that suggest that this behavior can be explained by densely packed asperities, which act to widen the length scale of the slip pulse, rather than a narrow slip pulse. In the third chapter, I explore small slow slip events (SSEs), with Mw < 6.0, and assesses whether fault slip and tremor detections scale linearly. Under the assumption that tremor and slip are spatially and temporally related during slow slip events, I develop a scaling relationship between tremor counts and slip based on known large slow slip events (SSEs) in Cascadia. I use the existing tremor catalog in Cascadia to cluster tremor detections into distinct events that can be scaled into slip distributions. Using this scaling relationship on a clustered tremor catalog, I obtain event magnitudes that range from Mw 4.5 to 6.5. We also find that the larger (Mo > 3*1017 Nm) clustered events follow a Mo ~ T scaling. This catalog partially fills the long-standing observational gap between seismically detectable events and geodetically detectable events. GPS and strainmeters are used as an independent check of the scaling relationship. Using this clustered catalog as a guide, I identify a patch of repeating events beneath the Olympic Peninsula that produces frequent small SSEs. We stack the daily GPS time series for seven small slow slip events, aligning each record on peak tremor activity. We then estimate the average surface displacement and find the average moment. The GPS-based average moment for events in this patch is Mw 5.5 with peak fault slip reaching 0.6 cm and only 30% of the tremoring area slipping, compared to Mw 5.8 predicted by scaling the number of tremor detections. For further validation of our scaling relationship, I compare the scaled-tremor models to observed strainmeter records. We find that our empirical scaling relationship for large SSEs accurately predicts the strain for several small SSEs.