United States Geological Survey Earthquake Hazards Program Final Technical Report USGS Award Numbers G17AP00038 and G17AP00039 Improved Characterization of Slow Slip in Cascadia by Stacking GPS on Tremor Times: Collaborative Research with University of Alaska-Fairbanks and Miami University Authors: Stephen Holtkamp Geophysical Institute University of Alaska Fairbanks 903 Koyukuk Dr., PO Box 757320 Fairbanks, AK, 99775 Office: (907)474-5751 [email protected] Michael Brudzinski Department of Geology and Environmental Earth Science Miami University 250 S. Patterson Ave. Miami University Oxford, OH 45056 Office: (513) 280-0660 [email protected] Other Personnel: Nealey Sims, UAF graduate student Shannon Fasola, Miami University graduate student Award Term: February 1 2017 – January 31 2018 Acknowledgement of Support and Disclaimer: This material is based upon work supported by the U.S. Geological Survey under Grant No. G16AP00140. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the opinions or policies of the U.S. Geological Survey. Mention of trade names or commercial products does not constitute their endorsement by the U.S. Geological Survey. Abstract The origins of slow slip and how transient slip processes factor into the nature of the subduction zone seismic cycle have been the focus of research since the discovery of slow slip in subduction zones. Such efforts have included, but are not limited to, (1) understanding the along-dip extent of slow slip events, with one goal of this being to determine how slow slip events may relieve or increase stress on the locked portions of the megathrust (Beroza & Ide, 2011); and (2) probing the underlying fault physics for this new class of slow earthquakes, for example by determining various scaling relations (Aguiar et al., 2009). Despite these efforts, many key observations could be improved upon or are poorly resolved, such as the updip extent of slip (the vertical component of geodetic observations is often ignored because of high noise levels), or the nature of slip during inter-ETS events. In this study, we have investigated the viability of improving our characterization of slow slip to answer the questions posed above. We focused our efforts on two techniques to stack GPS time series (for each station individually, not cross-station stacking) using times of tremor episodes to produce geodetic observations of slow slip which have significantly higher signal to noise ratios. First, we utilized catalogs of individual tremor events to determine the timing of GPS stacks. Second, we utilized single seismic station mean amplitude processing to determine the timing of when to stack GPS signals. Both methods produced observations with increased signal to noise ratios that improve our ability to resolve slow slip. We have found that by stacking even a modest number of events (~10), we are able to resolve slips of ~0.5 mm. For ETS events, GPS stacking produces significantly better vertical observations and observations from stations up to ~200km away from the slipping patch. For the smaller inter-ETS events, which previously have only been observed geodetically in strainmeter time series, GPS stacking produces observations robust enough to invert for slip on the interface. Preliminary inversions of surface displacements to slip on the plate interface reveal that slow slip during inter-ETS times are deeper on the plate interface, consistent with the independent evidence from that inter-ETS tremor locations occur deeper as well. However, the location of inter-ETS slow slip is still shifted slightly updip from the zone of inter-ETS tremor, similar to what has been observed for ETS. Inversions of enhanced signals at the southern end of the Cascadia margin also indicate that some slow slip is occurring at the southern edge of the subducted slab, about 150 km south of the surface location of the Mendocino triple junction. 1. Introduction Fault displacement during large earthquakes is a consequence of interseismic locking, the degree and spatial variability of coupling between the two plates. The discovery (Dragert et al., 2001; Ozawa et al., 2001) and subsequently recognized ubiquity (Peng & Gomberg, 2010) of slow slip phenomena at subduction margins worldwide has obscured the nature of interseismic coupling along subduction margins, but at the same time has led to a proliferation of theories regarding potential interaction between slow slip and large megathrust earthquakes (Bouchon et al., 2011; Kato et al., 2016; Mavrommatis et al., 2015). Aside from any potential relation to seismogenic locking, observations of slow slip phenomena have greatly informed theoretical efforts to understand strain accumulation and release along the transition zone from seismogenic locking and extending into the stable sliding zone (Wech et al., 2009; Wech & Creager, 2011). A variety of seismic and geodetic signals comprise the continuum of slow slip phenomena. Seismic examples recorded at subduction margins include tectonic tremor (Obara, 2002; Rogers & Dragert, 2003), low frequency earthquakes (LFE’s) (Shelly et al., 2006), very low frequency earthquakes (VLFE’s) (Ghosh et al., 2015; Ito et al., 2007), and earthquake swarms (though not directly – aseismic slip is often inferred) (Holtkamp & Brudzinski, 2011; Lohman & McGuire, 2007). Geodetic manifestations of slow slip include fault afterslip (Heki et al., 1997), short- (Obara & Hirose, 2006), and long- term (Larson, 2004; Ohta et al., 2006) slow slip events (SSE’s), which are readily observed in GPS and/or strainmeter data. In Cascadia, slow slip events and tremor occur together and are called Episodic Tremor and Slip (ETS). They were coined “episodic” because (1) they can occur with a high degree of regularity, and (2) they represent repeated failure of the same source region along the subduction megathrust. While slow slip events have been observed updip from the locked zone at other convergent margins worldwide (i.e., in the forearc wedge), ETS events in Cascadia are observed down-dip from the locked zone. Geodetic studies of interseismic locking in Cascadia generally agree that the locked zone is separated from the tremor zone by a resolvable, sometimes significant gap (Schmalzle et al., 2014), with several researchers proposing a potential zone of persistent fault creep separating the locked and tremor regions (Hyndman, 2013). Alternatively, slip during ETS events could be propagating up-dip from the tremor source region, a hypothesis supported by some geodetic modeling (Wech et al., 2009). In any case, the region between the locked zone and the tremor region is crucial to understanding any potential relation between ETS and seismic hazard. Aside from hazard implications, the nature of ETS is now considered to be crucial to understanding fault physics and mechanics, particularly with respect to depth on the megathrust. Wech et al. (2011) show, using tremor observations, that slip size and periodicity vary considerably with depth. They find that tremor on the deepest portion of the tremor zone occurs as smaller, more frequent bursts, and that tremor on the shallowest portion of the tremor zone occurs as larger, less frequent bursts, with a continuum of behaviors in between. This implies that the fault is weakening with depth. ETS events begin with deep tremor before transitioning into shallower regions with geodetically observable slip, at which point it is considered to be an ETS event. The similarity between the sources of tremor and slip during ETS events (Bartlow et al., 2011) has led to the development of a scaling law (Aguiar et al., 2009) relating the moment magnitude of the slow slip to the total duration of tremor (e.g., number of hours tremor is present). Unfortunately, GPS displacements expected from inter-ETS events are below the noise level of modern GPS receivers (<1mm maximum expected displacement). Wech et al. (2011) has shown that inter-ETS and ETS tremor behave in a self-similar manner, but a lack of GPS observations from inter-ETS events has prevented comparisons to ETS slip, and has prevented the scaling law hypothesized by Aguiar et al. (2009) from being extended to small events. Our objective with this project was to produce high quality stacked GPS observations of both ETS and inter-ETS events in Cascadia, and invert those surface displacements to slip on the plate interface. Current geodetic transient detection methods operate on individual slow slip events in the geodetic time series (e.g., Schmalzle et al., 2014). These detection methods require that the signal from a slow slip event is greater than the noise level of the geodetic network. In the horizontal component, this threshold is typically greater than 1 mm, and in the vertical component can be several mm. In Cascadia, this means that slow slip events with cumulative magnitude less than about Mw = 6 do not generate displacements at the surface which are robust. Also, most studies of slow slip events in Cascadia have not utilized vertical observations of slip, even though this component is most sensitive to the depth distribution of slip. Through GPS stacking (Frank et al., 2015), we first show that we can achieve significantly greater signal to noise ratio observations of ETS events along the Cascadia margin. We focus on two regions: Puget Sound and the Mendocino Triple Junction. We chose Puget Sound as our first focus area because this region has been extensively benchmarked by prior geodetic studies (Bartlow et al., 2011; Dragert et al., 2001; Holtkamp & Brudzinski, 2010). We show that we are able to resolve the horizontal deformation field out to hundreds of km away from the source region. We are also able to resolve the vertical deformation field throughout the region of interest, a key advancement from prior efforts. We next focus on the Mendocino region because very few studies have been able to extract robust estimates of slow slip in this tectonically complex region. We show that we are able to resolve observations of slow slip well south of the Mendocino Triple Junction, in contrast with prior studies (Schmalzle et al., 2014; Szeliga, et al., 2004).