IMAGING THE SLAB AND FOREARC ABOVE THE 2010 MAULE EARTHQUAKE IN CENTRAL CHILE by Mallory Morell A Prepublication Manuscript Submitted to the Faculty of the DEPARTMENT OF GEOSCIENCES In Partial Fulfillment of the Requirements for the Degree of MASTER OF SCIENCE In the Graduate College THE UNIVERSITY OF ARIZONA 2012 1 Imaging the Slab and Forearc Above the 2010 Maule Earthquake in Central Chile Mallory Morell, Susan Beck, George Zandt Abstract The Mw=8.8 Maule earthquake that occurred off the coast of Chile on February 27, 2010 is one of the largest megathrust earthquakes ever to be recorded and ruptured ~600 km of the plate boundary. This segment of the Nazca-South America plate boundary is an ideal region to investigate the processes related to the structure of the forearc wedge, continental Moho and subducting slab because of all the seismic data recorded following the Maule earthquake. Within weeks of the Maule earthquake, international teams joined Chilean seismologists to install an array of seismic stations between 33°- 38.5ºS, from the coast to the foothills of the Andes to produce the International Maule Aftershock Deployment (IMAD) data set. These arrays were deployed ~6-10 months in order to capture and study the aftershocks in and around the rupture zone, and to better understand crustal and mantle wedge structure. We calculated receiver functions (RFs) from P and PP phases and made Common Conversion Point (CCP) stacks to image the structures in the slab and forearc wedge. In our trench parallel cross-sections near the Chile coast we image a strong P-to-S conversion we interpret as the down-going oceanic Moho. In addition we see a arrival indicating a decrease in velocity that we interpret as the base of the slab (oceanic LAB) indicating the oceanic slab is ~50-60 km thick. Near the hypocenter of the Maule earthquake the slab arrival is disrupted and is interpreted to be faulted or imbricated at a depth of ~40 km. In our trench normal sections we see the oceanic Moho to a depth of ~70-80 km and the continental Moho at ~35 km in the forearc where it approaches the slab. Finally, we combined IMAD data with previously collected CHARGE data to create a more complete CCP RF cross-section across the entire Andes at ~36ºS. This is the first study that has imaged the continental Moho in the forearc at this latitude (36º-37.5ºS) or to use RFs to image the rupture features from the Maule event. Introduction The Chilean subduction zone is one of the most active convergent plate boundaries on Earth, producing large subduction zone megathrust events every couple of decades. Some examples include the Mw=9.5 1960 event that was the largest instrumentally observed 2 earthquake to date and devastated the Pacific rim with its resulting tsunami [Cifuentes, 1989; Barrientos and Ward, 1990] and the more recent 2010 Mw=8.8 Maule earthquake that created a local tsunami and whose shaking affected as many as 1.8 million people [USGS, 2012]. This subduction system is also responsible for creating some of the tallest and most active volcanoes on Earth, including the stratovolcano Cerro Azul whose peak is 3788 m high and was the source of the world’s largest explosive eruption of the 20th century in 1932 that ejected 9.5 cu km of dacitic tephra into the atmosphere [Siebert and Simkin, 2002-2012]. The Mw=8.8 Maule earthquake ruptured ~600 km of the subduction zone on the 27th of February 2010 causing a damaging tsunami locally and 500 deaths [USGS, 2012]. Immediately after this megathrust earthquake, a large international response was organized to deploy seismic instruments in the aftershock region. The International Maule Aftershock Deployment (IMAD), a joint project between the University of Chile, IRIS (Incorporated Research Institutions for Seismology), IPGP (Institut de Physique du Globe de Paris), Caltech, GFZ (Deutsches GeoForchungsZentrum), and the University of Liverpool, was comprised of 104 broadband stations located at 99 sites deployed two weeks to 1 month after the mainshock with the goal of recording aftershocks (Fig. 1). The best station coverage was achieved between April 1 and June 1, 2010 and the data were made available as soon as it could be archived. The station spacing across this array was on average 30 km and covered the whole length of the rupture from 32.5º38.5ºS. Lange et al., [2012] and Rietbrock et al., [2012] have determined 20,205 and 30,000+ aftershock locations, respectively, using data collected by this international array. In addition, Hayes et al. [in preparation, 2012b] have also relocated Maule earthquake aftershocks using the IMAD data. All three studies show similar patterns with the deepest aftershocks occurring at ~50-55 km depth along the plate interface and a noticeable gap in aftershocks between ~35-40 km depth. Our goal in this study was to image the subducting slab, forearc, and arc, and examine how the subducting slab interacts with the continental Moho in the forearc using receiver function analysis. We found the location and density of this array to be optimal for expanding upon past receiver function studies that had also incorporated the southern line of the CHARGE array [Fig. 1; Gilbert et al., 2006] and to compare with local source tomography studies conducted over the 1960 rupture further south as discussed below [e.g. Haberland et al., 2006, 2009; Boehm et al., 2002; Yuan et al., 2006; Asch et al., 2006]. 3 Tectonic Setting Between 32º and 39ºS along the south central Chilean continental margin, the Nazca plate is converging obliquely with the South American plate at a rate of ~7.4 cm/yr with an azimuth of ~74º [DeMets et al., 2010]. This angle of convergence has caused a series of along-strike fault systems, including the 1100 km long right-lateral Liquiñe-Ofqui Fault Zone that follows the foothills of the Main Cordillera or active arc [Haberland et al., 2006]. The age of the subducting plate in this region varies between 30-35 Ma, with age increasing to the north [Müller et al., 1997]. The Juan Fernandez Ridge and its resultant flat slab region bound the 2010 Maule rupture zone to the north and the Mocha, Valdivia, and Bio-Bio Fracture zones act as a southern boundary. The southern boundary is also where the 1960 Mw=9.5 rupture nucleated and where we find the anomalous Arauco Peninsula [Hicks et al., 2012]. The overriding plate above the 2010, Maule Earthquake rupture can be divided into three trench-parallel geomorphic segments – the Coastal Cordillera, the Central Valley, and the Main Cordillera or active arc. Where our cross sections intersect the active arc and extend into the backarc, they cross a pair of historically active volcanoes, Descabezado Grande and Cerro Azul [Siebert and Simkin, 2002-2012, Fig. 1]. Previous Geophysical Studies The region between 32º-39ºS has been targeted in the past for geophysical studies in part because of the transition from flat slab subduction to the north back to normal subduction, the location of the 1960 epicenter just south of the region, and the gap in seismicity that has been observed between 35ºS-37ºS [Campos et al., 2002]. There have been several arrays deployed across Chile and western Argentina (e.g. CHARGE 2000-2002; ISSA2000 2000; TIPTEQ 20042005; temporary seismic installations between 39ºS-40ºS 2008-2009). The 8 stations in the southern transect of the Chile Argentina Geophysical Experiment (CHARGE) array were the only stations formerly deployed for any length of time within our study area. These stations were spaced an average of 70 km apart, which allowed for the imaging of large scale structures such as the continental Moho and crustal structures beneath the arc, but had limited resolution in the forearc [Gilbert et al., 2006; Wagner et al., 2005; Heit et al., 2008]. A transect through the center of our study region at 36ºS was imaged using P-wave receiver functions (RFs) and S-wave receiver functions (SRFs) by Gilbert et al. [2006] and Heit 4 et al. [2008], respectively. They observed the continental Moho beneath the arc and in the backarc at depths between 25-55 km. Heit et al. [2008] were also able to image the LAB (Lithosphere-Asthenosphere Boundary) beneath the continental Moho in the backarc at depths between 100-125 km. Neither study had resolution in the forearc, so neither was able to robustly image the downgoing slab in this region. Using local travel time tomography, Wagner et al. [2005] found low vp, low vs, and a high vp/vs directly beneath the arc at a depth of 50-100 km in this region that was interpreted as localized pockets of melt. South of the Maule aftershock region, at ~39ºS, receiver function studies by Asch et al. [2006] and Yuan et al. [2006] and local travel time tomography studies by Haberland et al. [2009] and Boehm et al. [2002] imaged the downgoing slab in great detail. Both receiver function studies imaged the slab down to a depth of 110 km, closely following local seismicity. They also observed the continental Moho reaching its maximum depth of 40-50 km beneath the arc and shallowing to 25 km toward the west and 35 km toward the east. The local source tomography studies observed the downgoing slab to a depth of 120 km, dipping 30ºE, and they also observed a thickening of the continental crust under the arc to ~50 km and thinning to the west. Haberland et al. [2009] noted that the mantle wedge is most likely not serpentinized throughout, but instead is hydrated in patches. What is lacking for these studies is a clear image of the continental Moho in the forearc and at 36ºS, a clear image of the downgoing slab. Data and Methods Data IMAD stations contained 4 different sensor types: Guralp CMG-3ESP Seismometer (30s – 50Hz), Nanometric Trillian 120P (120s – 10Hz), Guralp CMG40 (1Hz – 100Hz), and Guralp CMG3T (120s – 50Hz). In addition to the IMAD experiment, we incorporated data from the CHARGE portable broadband seismic experiment deployed in 2000-2002 as well as permanent stations PEL from the Geoscope Global Network and GO05 from the Chilean National Network (Fig. 1). This IMAD deployment was designed with the primary goal of recording aftershocks following the Mw=8.8 Maule earthquake. This made identifying teleseismic events challenging because the aftershocks created “noise” that had to be filtered out, and in some cases made teleseismic P- or PP waveforms unusable. P phases were chosen for large magnitude (Mw > 6) 5 earthquakes at distances between 30º-95º and PP phases (Mw > 6.5) from 95º-180º. We were able to use four events with well-recorded P phases and five events with well-recorded PP phases. This resulted in high quality receiver functions, though it also limited our backazimuth coverage. For the longer deployed CHARGE array we were able to use 66 events for P phases. Methods We carried out RF analysis on teleseismic P and PP phases using data recorded during the IMAD experiment and data from other broadband stations in the region. Initially, seismograms were bandpass filtered from 0.05 to 3 Hz, to help remove the thousands of local aftershock events and anthropogenic noise. Then the horizontal components were rotated to the radial and tangential components. For the RF calculations we used an iterative time domain (pulse stripping) deconvolution approach [Ligorria and Ammon, 1999], in which the vertical component is deconvolved from the horizontal components in a 110 s time window, starting 10 s before the P or PP onset. This process forms a series of Gaussian spikes that include P-S conversions from discontinuities. Within the RF the amplitude and timing of the P-SV conversions can be directly related to the discontinuity’s impedance contrast and depth below the station [e.g. Langston, 1977; Owens et al., 1984]. We tested two different Gaussian values, 1 and 2.5, which act as low pass filters, (this corresponds to low-pass filter corner frequencies of 0.5 and 1.2 Hz respectively). For the results presented here, we show a Gaussian of 2.5; making our maximum resolution layers > 1 km. We calculated unnormalized RF and all of the radial RFs were required to have a variance reduction of at least 70%, meaning 70% of the signal is fit. Next, the RFs were visually inspected to eliminate those of poor quality. Some qualities used to determine if a trace should be excluded included a negative first arrival, large delay in the first arrival, and extreme ringing along the trace. Of the 781 starting RFs for a Gaussian of 2.5, 376 unnormalized remained after this quality control. Once the RFs were calculated, we generated Common Conversion Point (CCP) receiver function cross-sections migrated to depth [Dueker and Sheehan 1997; Gilbert and Sheehan 2004]. Stacking numerous RFs from various stations helped minimize the noise caused by near surface effects that are specific to each station and not coherent between stations. Over 370 RF traces (with a Gaussian of 2.5) were stacked to create our final cross-sections. The P and 6 converted S waves are ray-traced through geographic bins using a constant vp and vp/vs. The rays that went into each stack were determined by the bin spacing that in our case varied between 5 km and 15 km. We used a bin sharing of 2, meaning that the data from one bin in each direction was averaged with the data from the center bin to calculate each stack. This helped eliminate gaps in our cross-sections but also caused some lateral smearing. The exception to this bin sharing value was Profile X2 where a bin sharing of 1.5 was used. The average number of good RFs at a station was 2 with the maximum of 7 at L006 and U59B, and a minimum of 1 RF at 20 stations. Of the 104 stations installed, we were able to use data from 87 stations. At the permanent station PEL we used 40 events and at GO05 we used 4 events. Despite the limited number of earthquakes we have a total of 109 stations with high quality P and PP phases recorded. For this study several different cross-sections and bin spacings were used: 1) for the trench-parallel cross-sections (Profiles Y1-4 on Fig. 1), the bin centers are separated by 15 km in both the NE-SW and NW-SE directions and the mesh was rotated 15º clockwise from north, 2) trench perpendicular Profile X3 the bin centers are separated by 15 km in both the NE-SW and NW-SE directions and the mesh rotated 15º counterclockwise from east, 3) trench perpendicular Profile X2 differs only by having a bin spacing of 5 km in the NE-SW direction and 20 km in the NW-SE directions, 4) for trench perpendicular Profile X1 the bin centers are also separated by 15 km in the N-S and E-W directions and the grid is not rotated, and 5) for Profile XCHRG the bin centers are separated by 20 km in the N-S direction and 25 km in the E-W direction, have a bin sharing of 2, and are also rotated 15º counterclockwise from east (Fig. 1). To migrate from time to depth in the CCP cross-sections we used a vp = 6.5 km s-1 above 60 km, to insure crustal features are not mapped artificially too deep by migrating them with mantle velocities, and vp = 8.0 km s-1 below that and a vp/vs of 1.75. This vp/vs value is the same as assumed by Campos et al., [2002] in their earthquake location study across the region and is the average value of the gradient calculated by Haberland et al., [2006, 2009] for the segment of the subduction zone just south of our region. Changing the value assigned to the vp/vs does not change the pattern within the CCP stacks, it would simply shift the depth of the discontinuities slightly (Fig. S1). Amplitudes were rescaled to take into account the varying angles of incidence of the earthquake arrivals by multiplying each trace by ir/i, where ir is a depth invariant reference angle (20º) and i is the incident angle calculated from P slowness [Jones and Phinney, 1998]. 7 Modeling RF at Station PEL As noted in many previous receiver function studies conducted in central and southern Chile, most of the teleseismic earthquakes occur in Pacific rim subduction zones, making it difficult to have the full range of backazimuths that would be ideal for this type of study [e.g. Heit et al., 2008; Gilbert et al., 2006; Yuan et al., 2006]. South central Chile has only one active subduction zone located to its east-southeast, the South Sandwich Islands, that were not active with large events (>6.0) during the 10 months the stations were deployed. The earthquakes used in our final dataset are located mainly to the west (Tonga Trench) and the northwest (Baja California and Ecuador) of our region, with very few events from the east or southeast (Mid Atlantic ridge, South Sandwich Islands). Events occurring at eastern azimuths are key to imaging a slab dipping to the east, as indicated by studies done by Owens et al [1984] which illustrated that waves incident from western azimuths for an eastward dipping slab produce the smallest amplitude converted phases. In order to look in more detail at RFs from a broad range of backazimuths, we incorporated data recorded at PEL, a permanent station from the Geoscope Global Network. Station PEL has been operating since 1998, hence we identified 40 wellrecorded teleseismic earthquakes for a more detailed analysis. We calculated receiver functions for PEL and then calculated synthetic RFs for different velocity structures to determine a more detailed structure than we could obtain from the data recorded at the temporary stations. Figure 2 shows the RFs plotted as a function of backazimuth that shows a complex pattern of converted phases. We use this suite of RFs and forward modeling to look at the sensitivity of the RFs to velocity models with dipping layers (slab and Moho). We calculated synthetic teleseismic waveforms with a full range of backazimuths starting with a basic 3-layer over half space seismic velocity model containing continental crust, continental Moho, oceanic slab, and oceanic Moho, using the program Ray3D [Owens et al., 1984]. We then calculated RFs from the synthetic waveforms using the same iterative time domain deconvolution approach with the same Gaussian value of 2.5 that was applied to the data recorded at PEL. The RFs for both PEL and the synthetics were then binned and plotted by backazimuth. Following the approach by Cassidy [1992], we started with a basic 3-layers over half space model in order to constrain the orientation of large-scale features. As this region is much more complex than a simple 3-layers over half space model, a more complex model was needed to explain more of the prominent inter-crustal arrivals. We added inter-crustral layers one 8 at a time and adjusted each layer’s depth until we developed a model that approximately fit the RFs at PEL. We identified 2 inter-crustral layers above the continental Moho. This gave us a final model with 5-layers over half space (Fig. 3). Results We first describe our results from the permanent station PEL and then go on to describe our results in the forearc and across the arc in south central Chile. Station PEL Although forward modeling of RFs is non-unique, we can use RFs recorded at PEL to test the sensitivity of dipping layers and event backazimuth. Figure 3 compares plots of the synthetic RF traces distributed by backazimuth created from the final 5-layer model with RFs calculated from data recorded at PEL. Although not every detail fits, the first order patterns appear very similar. The 5-layer over half space model was built based on previous work in the region. The slab contours (slab1.0) from Hayes et al. [2012a] indicate that the top of the slab is at ~70 km depth under PEL and these slab contours are consistent with oceanic Moho depths estimated at 2 other IMAD stations (Fig. 4). We estimated the oceanic crustal thickness to be 10 km based on typical oceanic crustal thicknesses [Leeds et al., 1974] and the RF sections we have stacked. We used a depth to the continental Moho of 42 km. Using these depths and thicknesses, we created the initial 3-layer model. To define the intercrustal arrivals we observe in the PEL RFs we looked to the topography and geology. There is a prominent positive arrival following the direct P-arrival, which is followed by a very prominent negative arrival. The prominent positive arrival might be explained by PEL’s location. Since PEL sits in the foothills of the Andes, it is possible that there is a thin high-velocity impedance contrast dipping to the west, produced by the base of a forearc sedimentary basin. The negative arrival that follows is very prominent for a wide range of backazimuths, indicating it is relatively flat-laying. In our updated model we added two inter-crustal discontinuities as shown in Figure 3b. Using our new 5-layer model our synthetic RFs have similar patterns to the observed RF from PEL. We can make several important observations from these results. The first observation is that the slab arrival (top of the oceanic crust and oceanic Moho) are strongest at back azimuths between 0 and 160° (northeast-east-southeast) where we have limited data recorded at the 9 temporary stations. More earthquakes from the east would improve our image of the eastdipping slab. Second, many of our events are from backazimuth directions (200 to 350°) where the signal from the slab disappears due to the angle of incidence between the P or PP phase with the slab. Hence, our image of the slab in cross-sections perpendicular to the trench is limited in resolution. The same is true for an east dipping continental Moho, the strongest conversions are from stations between 0 and 180° and not from backazimuths of 200° to 350° Chile Forearc CCP RF stacks from the forearc region are shown in Figures 5 and 6. We show sections both parallel (Fig. 5) and perpendicular (Fig. 6) to the trench in regions where we have the most data to stack and can produce near complete cross-sections. We show four trench-parallel CCP cross-sections in Figure 5 for the line locations shown in Figure 1. The most prominent P-to-S conversion shows up as an increase in S-wave velocity with depth (red arrival) seen as a continuous horizontal feature with a depth varying between 3560 km (Fig. 5). It also closely follows the seismicity, which clusters along a narrow band. We interpret this feature as the oceanic Moho in the subducting slab. We would expect the oceanic Moho to be an increase in velocity between the oceanic crust and mantle and the depth of the conversions is at the right depth based on seismicity. In Figure 5 the white circles are Maule aftershock events relocated by Hayes et al. [in preparation, 2012b] and the black circles are events from 1970 to 2012 relocated by the NEIC [USGS, 2012]. The dashed black line indicates the location of the slab calculated from the USGS Slab1.0 model [Hayes et al., 2012a] and the white dashed line indicates our estimate of the location of the LAB (Fig. 5). At a depth between 95-110 km there is another prominent horizontal conversion that indicates a decrease in velocity with depth (blue) that we interpret as the oceanic lithosphereasthenosphere boundary (LAB). Based on the timing we conclude this conversion is a primary arrival and not the second multiple of the oceanic Moho. This gives us a thickness of the oceanic lithosphere of ~50-60 km which fits with the predicted thickness based on the 35-40 Ma age of the subducted lithosphere [Leeds et al., 1974]. The subducting oceanic Moho varies slightly as you move from north to south along the trench parallel sections (Fig. 5). Profile Y1-Y4 in Figure 5 runs sub-parallel to the down-dip edge of the Maule earthquake rupture zone and profile Y1 likely includes the hypocenter where 10 the mainshock rupture initiated (white star in Figure 5). At ~35.6º-36°S and at a depth of 35-40 km, there is a disruption in the positive arrival that we interpret as the oceanic Moho that can be most prominently seen in Line Y1, but also appears in profiles Y2 and Y3. The disruption in Y1 looks as if the oceanic Moho has been uplifted ~10 km higher than the surrounding oceanic Moho arrival. This same feature can be seen in profiles Y2 and Y3 at the same location and depth, but instead of an offset of the oceanic Moho, it appears as an arrival above the Moho. At the north end of our cross-sections (33.5°-34.5°S) there are positive arrivals directly beneath the oceanic Moho. This would appear to be a discontinuity within the down going slab. Along parts of our cross-sections there is a negative arrival above the positive oceanic Moho arrival that might be the top of the oceanic crust suggesting in places that the thickness of the oceanic crust is ~10 km depth. This negative arrival is discontinuous, hence we can’t be sure if it consistently represents the top of the oceanic crust. At a depth of 40-50 km in the subduction system it is not clear if we should see a decrease in seismic velocity at the top of the oceanic crust since we are not sure what the material is above the slab. If the material above the slab is continental crustal material then the oceanic crust would represent an increase in velocity but if there is continental mantle material above the oceanic crust then we would expect to see a decrease in velocity. If the material above the slab is hydrated mantle then we would not expect to see the oceanic crust. Along profile Y4, the slab shallows to the north, hence, we expect to see the oceanic Moho arrival at a shallower depth than further south. In profile Y4 the oceanic Moho arrival is more discontinuous than the other cross-sections and appears to have an abrupt change in depth between 35º-35.5ºS consistent with the general trend of the slab contours from the USGS [slab1.0 model, Hayes et al., 2012a]. We cannot be sure if this change in depth is as abrupt as it appears or if we lack the data coverage to see a more gradational change. Our most complete trench normal section is profile X2 in Figure 6. We identify the oceanic Moho as a series of red arrivals that dip down to ~80 km. In general, we do not see the top of the oceanic crust along the dipping slab in part because it may not have a distinct velocity contrast to the over-riding forearc and or mantle material. There is a negative conversion at ~20 km just as the slab begins to descend into the mantle that is a possible candidate for the top of the oceanic crust. The seismicity follows the prominent positive arrivals down to a depth of ~80 km. The downgoing slab does not appear very continuous, which is likely due to the limited 11 backazimuth directions of the events as was discussed in our synthetic modeling section. At the bin spacing used (5 km), there is still some lateral streaking, making the downgoing plate appear segmented. In the X2 profile, there is also a prominent positive arrival at ~35-40 km depth that we interpret as the continental Moho under the foothills of the Andes. This discontinuity does not reach the slab but the projection is consistent with the down dip limit of the Maule earthquake rupture zone near a depth of 40 km [Lorito et al., 2011, Vigny et al., 2011]. The trench perpendicular profiles X1 and X3 have less data so we used a bin spacing of 15 km and bin sharing of 2, which results in more lateral smearing, hence we do not see a clear indication of the slab, however we do see the continental Moho and mid-crustal structure. The line further to the north, profile X3, cuts far enough east to actually image below part of the active arc, which includes a historically active volcanic complex with two stratovolcanoes (Cerro Azul and Descabezado Grande). This line is just south of the CHARGE line that we will discuss below, which cuts across the active arc into the backarc. Profile X3 allows us to image the forearc of the same region in much greater detail. The most striking features in profile X3 are the two strong positive arrivals under the arc and foothills at ~30 and ~50 km depths (Fig. 6). Under the forearc we interpret a weaker positive arrival as the continental Moho at ~35 km depth. The CCP RF stacks from the southern forearc, between 37º-38ºS, are shown in Profile X1 where we used an east-west orientation, with a 15 km bin spacing and bin sharing of 2. Hence we see hints of the slab but there is significant lateral smearing. We observe a strong positive arrival under the eastern part of the forearc that we interpret as the continental Moho at ~35 km depth. Continental Moho across the Arc and Forearc We have combined RFs from the IMAD data with CHARGE data to calculate a more complete CCP RF stack across the forearc, the active arc, and into the backarc as shown in Figure 7. The most prominent feature on the cross-section is a strong increase in velocity in the backarc interpreted as the continental Moho. This strong P-to-S conversion arrival was imaged and identified as the Moho by Gilbert et al. [2006]. The continental Moho appears to bifurcate under the arc and appears as a series of discontinuities suggesting a layered structure under the arc between longitudes of 70° and 71°W. There is a strong conversion at ~30 km depth and a deeper conversion at ~60 km. We do not image the down-going slab in part due to the bin 12 sharing of 20 km, which would smear out dipping arrivals. There is no evidence in our sections for the boundary with the continental LAB at a depth of ~100 km as described by Heit et al. [2008] using S-wave RF. Discussion We have imaged the forearc between latitudes 32ºS and 39ºS using a relatively dense array of IMAD broadband stations. Two features that have been identified through this RF study include imaging of the subducting slab (interpreted as the oceanic Moho) and the continental Moho in the upper plate in the forearc and arc. In this section, we discuss our results and compare our interpretations with other studies in this region. Imaging the subducting slab We have imaged a P-to-S conversion in the depth range of 35-60 km in our trench parallel cross-sections that is at a depth consistent with the down-going slab (Fig. 5). Our interpretation is that it is most likely the oceanic Moho. In the discussion below we give two additional possibilities and argue that the arrival is most likely the oceanic Moho. There are 3 possible interpretations for the strong arrival we observed on the trench parallel cross-sections (Fig. 5). 1. Top of the oceanic crust. This interpretation would require low velocity continental crust above the slab. Hence, there would be an increase in velocity from continental crust (6.46.9 km/s) to oceanic crust (7.2-7.5 km/s). 2. Oceanic crustal Moho. In this case the material above the oceanic crust has P-wave velocities of in the range of 6.0-7.4 km/s consistent with some volume of hydrated and serpentinized material and would not produce a strong increase in velocity at the top of the oceanic crust. Hence, we are observing an increase from the oceanic crust (7.2-7.6 km/s) to the oceanic mantle (8.0-8.3 km/s). 3. Base of the hydrated oceanic mantle within the slab. If the down-going slab is hydrated into the upper mantle we might be seeing the increase in P-wave velocity from the hydrated mantle (7.0-7.5 km/s) to the normal mantle (8.0-8.3 km/s). Refraction studies done at 37-39°S, at the very end of the Maule rupture and further south in the SPOC project give us some insight into the velocity structure [Sick et al., 2006]. At 38.25°S the SPOC south line show P-wave velocities in the range of 6.8-7.0 km/s above the oceanic crust 13 and similar velocities in the down-going oceanic crust suggesting that we should not see a conversion at the top of the oceanic crust in the depth range of 25-45 km [Sick et al., 2006]. At 39°S seismic refraction data from the ISSA project show P-wave velocities above the oceanic crust ~7.2-7.3 km/s with the oceanic crust having a slightly lower velocity of 6.9-7.0 km/s in the depth range of 40-60 km/s [Boehm et al., 2002]. Haberland et al., [2009] also image the region between 37-39°S using travel time tomography and identify the oceanic crust as a high vp/vs layer and observe very little change in the P-wave velocity between the oceanic crust and the upper plate between 25-45 km depth with the largest P-wave velocity increase at the oceanic Moho. Hicks et al. [2012] using Maule earthquake aftershocks and local travel time tomography show a strong increase in P-wave velocity right below most of the aftershocks with velocities of 6.5-7.0 km/s above the deepest aftershocks at 30-50 km depth. Hicks et al. [2012] also see an increase in vp/vs in the region of the deepest aftershocks. However, they do not image the oceanic crust as a discrete layer, in part because they are using aftershocks, which may in large part represent slip and deformation at the plate interface, and hence may not be ideal in imaging the oceanic slab below the aftershocks. Clearly the region at the down-dip edge of a major earthquake rupture zone (35-60 km) is complex and processes such as fluid release are likely altering the material in the region. Based on the studies sited above and extrapolating north we suggest that the strong arrival in our trench parallel cross-sections is most likely the oceanic Moho and corresponds to case 2 above. Clearly more work is needed to refine the details of this critical region of the subduction zone (Fig. 5). Transect Across the Arc The southern transect of the CHARGE array cuts through the center of our IMAD array at 36ºS. The resolution in the forearc for previous RF studies conducted across this line had sparse station coverage and hence, limited resolution [e.g. Gilbert et al., 2006; Heit et al., 2008]. Our higher density array of stations focused over the forearc allows us to extend upon the observations made in previous studies. Gilbert et al. [2006] successfully imaged the continental Moho under the backarc using RFs. We were also able to clearly image the continental Moho in the backarc at ~40 km and the two positive arrivals at ~25-30 km and ~50 km depth under the arc similar to Gilbert et al. [2006]. Gilbert et al. [2006] interpreted the two arrivals under the arc 14 as different parts of a magmatic system, with the shallower discontinuity marking the location of the top of the lower crust and the deeper discontinuity is most likely the continental Moho under the arc. A strong negative arrival in the RF occurs between these positive arrivals and may indicate a region where basaltic material intrudes into the lower crust (Gilbert et al., 2006). While we cannot be sure of the actual structure under the active arc it is clear that there is a layered structure that does not extend all the way into the forearc or backarc, hence it is likely due to magmatic processes. As noted the section samples a very active portion of the arc where Cerro Azul had a major eruption in 1932. Gilbert et al. [2006] also described a large coherent negative arrival beneath the crust that they interpreted as the location of low-velocity material in the mantle that might be feeding the shallower crustal magma chamber. In their cross-section this feature is very prominent and continuous, across the entire section; within our section the negative arrival beneath the continental Moho is still continuous beneath the backarc but with a much lower amplitude. The negative arrival breaks up under the arc and ends as it approaches the location of the downgoing slab at a depth of ~70 km. The S-wave RF study by Heit et al. [2008] imaged a single continental Moho, with depths shallowing to nearly 25 km in the eastern part of the backarc and reaching its maximum thickness of 55 km under the arc. They also lose resolution under the volcanic arc, but see the Moho reappear in the forearc at ~35 km. If we assume the deeper positive arrival under the arc is the Moho, then these depths are consistent with the continental Moho in our sections, especially in the forearc. They also noted the presence of the continental LAB in the backarc at a depth of 100-125 km. However, there is no evidence of this feature in our CCP RF sections. Forearc – Southern lines The forearc between 38º-40ºS has been imaged in some detail using seismic data from the TIPTEC experiment [e.g. Haberland et al., 2006, 2009; Yuan et al., 2006]. Our southern-most profile, X1 (Fig. 1), falls between 37º-38ºS at the northern end of these previous studies. This region coincides most closely with the tomographic study by Haberland et al. [2009]. In the study by Haberland et al. [2009], they imaged “patches” of hydration in the forearc wedge. In our sections we see a large wedge of low-velocity between the slab and the crust. This could be consistent with hydrated patches being present; simply the difference between the hydrated 15 patches and the rest of the mantle wedge is not extreme enough to create visible boundaries in the RF stacks. The extent of the contact zone is highly debatable from our study and the RF studies by Yuan et al. [2006] and Asch et al. [2006]. They were unable to resolve the continental Moho above the forearc wedge, but estimate the depth could be as shallow as 25 km where it is cut by the Lanalhue fault zone. Our sections focus on the forearc, where we see the continental Moho approaching the slab at ~35 km, which is much shallower than those noted by Haberland et al. [2009] (~50 km) but deeper than the estimate made by Yuan et al. [2006]. Yuan et al. [2006] estimated that the continental Moho was raised due to movement of the Lanalhue fault system, but we do not see evidence for this because our cross-sections are too far north of the fault zone to observe such effects. Correlations with the 2010 Maule Rupture Zone Our coverage in the forearc is unique from previous studies in that most of our profiles, both along strike or normal to the trench, intersect with a portion of the rupture surface from the Maule mainshock especially as defined by the aftershocks. In the along strike sections we observe how the surface of the plate, in both the RF image and the USGS Slab1.0 model, displays a long wavelength undulation that may have influenced the location of the rupture interface. The most interesting correlation is that the rupture appears to have initiated where the slab interface is located at the lowest point of a long wavelength downward flexure (Fig. 5). We noticed in the along strike sections that where the rupture initiated there appears to be a discontinuity in the relatively continuous slab. This disruption appears as small block of uplifted slab that can be seen in the along strike sections further east as well, with the continuous slab appearing underneath the anomalous discontinuity (Y1 and Y2 in Fig. 5). A high-velocity feature was also imaged at approximately the same depth and location in the local source tomography study also using aftershocks recorded by the IMAD array [Hicks et al., 2012]. At 36ºS they noted a high vp (7.0 km/s) and a high vp/vs (~1.89) region at a depth of 25 km along the megathrust. They interpreted this as a topographic feature, possible subducted seamount on the down-going plate. We suggest that we are imaging the same feature in our trench-parallel sections (Fig. 5). We would expect a seamount to make the oceanic crust locally thicker, not necessarily cause an offset in the oceanic Moho. Rather, we interpret this feature to be faulting or 16 imbrications in the slab near the point of maximum downward flexure of the slab in the along strike direction, where along strike compression would be at a maximum in the upper part of the slab. We further suggest that this feature might have acted as a local asperity that served as a nucleation point for the Maule earthquake. Conclusions In this study, we used P and PP RFs to image the downgoing Nazca slab and forearc of the south central Chilean subduction zone above the 2010 Mw=8.8 Maule rupture zone. In addition, we extended the previous RF study across the entire arc and into the backarc. Our major conclusions are as follows: • We imaged the downgoing oceanic lithospheric Moho on trench parallel (near the coast) CCP RF cross-sections at 40-50 km depth. Above the oceanic Moho arrival is a discontinuous negative arrival that likely corresponds to the top of the oceanic crust indicating the crust is ~10 km thick, where it is present. • Our image of the subducted slab is disrupted by a section appear uplifted near the hypocenter of the Maule earthquake at a depth of ~40 km. This may indicate that the slab is faulted or imbricated in the region where the Maule earthquake initiated. • We image the base of the oceanic lithosphere as a decrease in seismic velocity indicating that the lithospheric thickness is 50-60 km, consistent with the age of the downgoing lithosphere. • Along trench perpendicular CCP RF cross-sections we see the oceanic Moho in the subducting slab down to depths of ~70-80 km. • We have a more complete CCP RF image across the entire Andes at ~36°S by combining IMAD data and previously collected CHARGE data. Similar to previous studies [Gilbert et. al., 2006] the crustal thickness is ~40 km in the backarc and increases to 60 km under the arc. The arc has a strong midcrustal discontinuity (increase in velocity) at ~30 km depth followed by a decrease in velocity above the Moho, yielding a strong layered structure in the lower crust that may be indicative of magmatic processes like basaltic underplating. • We observe the continental Moho at a depth of ~35 km in the forearc where it approaches the downgoing slab. These depths as well as the subducting slab dipping at ~35º are also 17 supported by our forward modeling under the permanent station PEL at the northern end of our array. This is the first study using CCP RF stacks that has imaged the continental Moho in the forearc at this latitude (36º-37.5ºS) or any of the rupture features from the Maule event. Future work is still needed to resolve how much of the forearc wedge is hydrated and to better constrain the uplifted section of the oceanic Moho seen near the point of nucleation of the Maule event. Acknowledgements This work was supported by NSF grant #1045597. Mallory Morell was also supported by scholarships from ConocoPhillips and ChevronTexaco. The U.S. IRIS Community response to collect IMAD data was made possible by NSF Rapid Response Research (RAPID) grant EAR1036349, EAR- 1036352 to the University of Florida and to IRIS respectively, and by the availability of EarthScope Flexible Array (FA) instruments. Special thanks to the GSAT group at the UA and all the IMAD field teams. References Asch, G., Schurr, B., Bohm, M., Yuan, X., Haberland, C., Heit, B., Kind, R., Woelbern, I., Bataille, K., Comte, D., Pardo, M., Viramonte, J., Rietbrock, A. & Giese, P., 2006. Seismological Studies of the Central and Southern Andes, The Andes, 443-457. Barrientos, S. & Ward, S., 1990. The 1960 Chile earthquake: Inversion for slip distribution from surface deformation, Geophys. J. Int., 103, 589-598. Boehm, M., Lüth, S., Echtler, H., Asch, G., Bataille, K., Bruhn, C., Rietbrock, A. & Wigger, P., 2002. The Southern Andes between 36º and 40ºS latitude: seismicity and average seismic velocities, Tectonophysics, 356, 275-289. Campos, J., Hatzfeld, D., Madariaga, R., Lopez, G., Kausel, E., Zollo, A., Iannacone, G., Fromm, R., Barrientos, S. & Lyon-Caen, H., 2002. A seismological study of the 1835 seismic gap in south-central Chile, Physics of the Earth and Planetary Interiors, 132, 177-195. Cassidy, J.F., 1992. Numerical Experiments in broadband receiver function analysis, Bull. Seism. Soc. Am., 82, 3, 1453-1474. Cifuentes, I.L., 1989. The 1960 Chilean Earthquake, J. Geophys. Res., 94, 665-680. DeMets, C., Gordon, R.G. & Argus, D.F., 2010. Geologically current plate motions, Geophys. J. Int., 181, 1-80. 18 Dueker, K.G. & Sheehan, A.F., 1997. Mantle discontinuity structure from mid-point stacks of converted P to S waves across the Yellowstone hotspot track, J. Geophys. Res., 102, 8313-8327. Gans, C.R., Beck, S.L., Zandt, G., Gilbert, H., Alvarado, P., Anderson, M. & Linkimer, L., 2011. Continental and oceanic crustal structure of the Pampean flat slab region, western Argentina, using receiver function analysis: new high-resolution results, Geophys. J. Int., 186, 45-58. Gilbert, H.J. & Sheehan, A.F., 2004. Images of crustal variations in the intermountain west, J. Geophys. Res., 109, B03306. Gilbert, H., Beck, S. & Zandt, G., 2006. Lithospheric and upper mantle structure of central Chile and Argentina, Geophys. J. Int., 165, 383-398. Groß, K., Micksch, U., and TIPTEQ Research Group, Seismic Team, 2008. The reflection seismic survey of project TIPTEQ—the inventory of the Chilean subduction zone at 38.2º S, Geophy. J. Int., 172, 565-571. Haberland, C., Rietbrock, A., Lange, D., Bataille, K. & Hofmann, S., 2006. Interaction between forearc and oceanic plate at the southcentral Chilean margin as seen in local seismic data, Geophys. Res. Lett., 33, L23302. Haberland, C., Rietbrock, A., Lange, D., Bataille, K. & Dahm, T., 2009. Structure of the seismogenic zone of the southcentral Chilean margin revealed by local earthquake traveltime tomography, J. Geophys. Res., 114, B01317. Hayes, G., Wald, D.J. & Johnson, R.L., 2012a. Slab1.0: A three-dimensional model of global subduction zone geometries, J. Geophys. Res.-Solid Earth, 117, B01302. Hayes, G., Bergman, E., Johnson, K., Benz, H., Brown, L., and Meltzer, A., 2012b. Seismotectonic framework of the February 27, 2010 Mw 8.8 Maule, Chile earthquake sequence, Geophys. J. Int.,in preparation. Heit, B., Yuan, X., Bianchi, M., Sodoudi, F. & Kind, R., 2008. Crustal thickness estimation beneath the southern central Andes at 30ºS and 36ºS from S wave receiver function analysis, Geophys. J. Int., 174, 249-254. Hicks, S.P., Rietbrock, A., Haberland, C.A., Ryder, I.M.A., Simons, M & Tassara, A., 2012. The 2010 Mw 8.8 Maule, Chile earthquake: Nucleation and rupture propagation controlled by a subducted topographic high, Geophys. Res. Lett., 39, L19308, 6, doi:10.1029/2012GL053184. Jones, C.H. & Phinney, R.A., 1998. Seismic structure of the lithosphere from teleseismic converted arrivals observed at small arrays in the southern Sierra Nevada and vicinity, California, J. Geophys. Res., 103, 10065-10090. 19 Lange, D., Tilmann, F., Barrientos, S.E., Contreras-Reyes, E., Methe, P., Moreno, M., Heit, B., Agurto, H., Bernard, P., Vilotte, J.P. & Beck, S., 2012. Aftershock seismicity of the 27 February 2010 Mw 8.8 Maule earthquake, Earth Planet. Sci. Lett., 413-425. Langston, C.A., 1977. The effect of planar dipping structure on source and receiver responses for constant ray parameter, Bull. seism. Soc. Am., 67, 1029-1050. Leeds, A.R., Knopoff, L. & Kausel, E.G., 1974. Variations of upper mantle structure under the Pacific Ocean, Science, 186, 141-143. Ligorria, J. and Ammon, C.J., 1999. Iterative deconvolution and receiver function estimation, Bull. seism. Soc. Am., 89, 1395-1400. Lorito, S., Romano, F., Atzori, S., Tong, X., Avallone, A., McCloskey, J., Cocco, M., Boschi, E. & Piatanesi, A., 2011. Limited overlap between the seismic gap and coseismic slip of the great 2010 Chile earthquake, Nature Geoscience, 4, 173-177. Melnick, D. & Echtler, H., 2006. Morphotectonic and geologic digital map compilations of the south-central Andes (36°-42°S). - In: Oncken, O.; Chong, G.; Franz, G.; Giese, P.; Götze, H.-J.; Ramos, V. A.; Strecker, M. R.; Wigger, P. (Eds.), The Andes - Active Subduction Orogeny., Springer. Müller, R., Roest, W., Royer, J., Gahagan, L. & Sclater, J., 1997. Digital isocrons of the world’s ocean floor. J. Geophys. Res., 102, 3211-3214. USGS, National Earthquake Information Center (NEIC): Earthquake Hazards Program, (http://earthquake.usgs.gov), November, 15th 2012. Owens, T.J., Zandt, G. & Taylor, S.R., 1984. Seismic evidence for an ancient rift beneath the Cumberland Plateau, Tennessee: a detailed analysis of broadband teleseismic P waveforms, J. geophy. Res., 89, 7783-7795. Rietbrock, A., Ryder, I., Hayes, G., Haberland, C., Comte, D., Roecker, S. & Lyon-Caen, H., 2012. Aftershock seismicity of the 2010 Maule Mw=8.8, Chile, earthquake: Correlation between co-seismic slip models and aftershock distribution?, Geophys. Res. Lett., 39, L08310. Sick, C., Yoon, M.K., Rauch, K., Buske, S., Lüth, S., Araneda, M., Bataille, K., Chong, G., Giese, P., Krawczyk, C., Mechie, J., Meyer, H., Oncken, O., Reichert, C., Schmitz, M., Shapiro, S., Stiller, M. & Wigger, P. 2006. Seismic images of accretive and erosive subduction zones from the Chilean margin, in The Andes – active subduction orogeny, eds Oncken, O., Chong, G., Franz, G., Giese, P., Götze, H.J., Ramos, V.A., Strecker, M.R. & Wigger, P., Frontiers in Earth Science Series, Vol 1. Springer-Verlag, Berlin Heidelberg New York, 147–170. Siebert L., & Simkin, T., 2002-. Volcanoes of the World: an Illustrated Catalog of Holocene Volcanoes and their Eruptions. Smithsonian Institution, Global Volcanism Program Digital Information Series, GVP-3, (http://www.volcano.si.edu/world/), November, 12th 2012. 20 Vigny, C., Socquet, A., Peyrat, S., Ruegg, J.C., Métois, M., Madariaga, R., Morvan, S., Lancieri, M., Lacassin, R., Campos, J., Carrizo, D., Bejar-Pizarro, M., Barrientos, S., Armijo, R., Aranda, C., Valderas-Bermejo, M.C., Ortega, I., Bondoux, F., Baize, S., Lyon-Caen, H., Pavez, A., Vilotte, J.P., Bevis, M., Brooks, B., Smalley, R., Parra, H., Baez, J.C., Blanco, M., Cimbaro, S. & Kendrick, E., 2011. The 2010 Mw 8.8 Maule Megathrust Earthquake of Central Chile, Monitored by GPS, Science, 332, 1417-1421. Wagner, L.S., Beck, S. & Zandt, G., 2005. Upper mantle structure in the south central Chilean subduction zone (30º to 36ºS), J. Geophys. Res., 110, B01308. Yuan, X., Asch, G., Bataille, K., Bock, G., Bohm, M., Echtler, H., Kind, R., Oncken, O. & Wölbern, I., 2006. Deep seismic images of the Southern Andes, in Evolution of the Andean Margin: A Tectonic and Magmatic View from the Andes to the Neuquén Basin (35º-39ºS lat), eds Mahlburg-Kay, S. & Ramos, V.A., Geological Society of America Special Paper, 407, 61-72. 21 Figure 1. Location map with IMAD broadband stations used in this study (squares) and cross section lines shown in Figures 5, 6 and 7. Two different slip models for the Mw=8.8 Maule event are outlined in blue and red. Coseismic slip model from Lorito et al. [2011] based on geodetic and tsunami data is shown in blue and the coseismic slip model from Vigny et al. [2011] based on GPS data is shown in red. Stations discussed in text are labeled in blue. Slab contours are the slab1.0 model from NEIC [Hayes et al., 2012a]. Volcano locations (red triangles) were taken from the Global Volcanism Program [Siebert and Simkins, 2002-2012]. 22 Figure 2. Receiver functions stacked by backazimuth plotted as a function of time for station PEL. Traces are sorted by backazimuth. Positive amplitudes are filled red and negative amplitudes are filled blue. 23 A ) B) Figure 3. Modeling results for PEL. (a) Synthetic (left) and calculated (right) unstacked radial and tangential receiver functions plotted as a function of time for station PEL. The synthetic receiver functions are calculated with the 5-layer block model. Traces are sorted by backazimuth and structures from the model are outlined and overlaid onto both the synthetic and calculated receiver functions (shaded regions). (b) Velocity model of the 5-layer model (left) and its representative block model showing east-west dips (right) of the major layers. 24 Figure 4. Stacked receiver functions from stations L006 and U59B, the stations with the most useable data within IMAD deployment. Thick red line indicates the positive arrival we interpret as the oceanic Moho below each station. 25 Figure 5. Common Conversion Point (CCP) cross-sections migrated to depth for trench parallel geometry. (left side) and our interpretation (right side). Note how the shape of the oceanic plate changes, becoming more “wavy” as the profiles move east away from the trench. Blue triangles are seismic stations, white star is Maule hypocenter taken from the NEIC [USGS, 2012], white circles are relocated aftershocks recorded during the IMAD experiment [Hayes et al., in preparation, 2012b], black circles are all earthquakes on record in the NEIC from 1970-2012 [USGS, 2012], black dashed line outlines the slab1.0 model [Hayes et al., 2012a], and white dashed line defines our estimate of the oceanic LAB. See Figure 1 for line locations. The cartoon of CCP Stacks shows our interpretation with the prominent red arrival the oceanic Moho. The black line is the slab1.0 model (Hayes et al., 2012a). 26 27 Figure 6. Common Conversion Point (CCP) cross-sections perpendicular to the trench (left side) and our interpretation (right side). Note how the shape of the forearc wedge changes. In profiles X1 and X2, there is a single continental Moho coming into a segmented slab, whereas in profile X3 the continental Moho bifurcates, making the forearc wedge appear much more complicated. Blue triangles are seismic stations, red triangles are volcanoes, white circles are relocated aftershocks recorded during the IMAD experiment [Hayes et al., in preparation, 2012b], black circles are all earthquakes on record in the NEIC from 1970-2012 [USGS, 2012], black dashed line outlines the slab1.0 model [Hayes et al., 2012a]. See Figure 1 for line locations. 28 Figure 7. Trench normal CCP receiver function cross-section (upper) following the CHARGE line and a cartoon of CCP cross-section (bottom) with our interpretation. Blue triangles are seismic stations, red triangles are volcanoes, white circles are relocated aftershocks recorded during the IMAD experiment [Hayes et al., in preparation, 2012b], black circles are all earthquakes on record in the NEIC from 19702012 [USGS, 2012] and the black dashed line is the slab1.0 model [Hayes et al., 2012a]. See Figure 1 for line locations. 29 Supplementary Tables/Figures Table S1. Earthquakes recorded by the IMAD array and used in this study Date Time Long Lat Depth Magnitude 4/4/10 22:40:43 -­‐115.278 32.297 4 7.2 4/6/10 22:15:01 97.048 2.383 31 7.8 6/12/10 19:26:50 91.936 7.881 35 7.5 6/16/10 3:16:27 136.543 -­‐2.174 18 7 7/23/10 23:15:10 123.259 6.776 640 7.5 8/12/10 11:54:15 -­‐77.306 -­‐1.266 206 7.1 9/3/10 16:35:47 171.83 -­‐43.522 12 7 10/21/10 17:53:13 -­‐109.159 24.69 10 6.7 30 Table S2. Earthquakes recorded by the CHARGE array and used in this study Date Time Lat Long Depth Mag 12/4/00 4:43:09 14.88 -­‐93.94 33 12/12/00 5:26:45 6.01 -­‐82.68 10 12/18/00 1:19:21 -­‐21.18 -­‐179.12 628 1/13/01 17:33:32 13.05 -­‐88.66 60 2/13/01 14:22:05 13.67 -­‐88.94 10 2/18/01 13:04:53 -­‐47.46 32.39 10 2/28/01 18:50:13 13.28 -­‐88.83 65 4/7/01 23:17:37 -­‐27.55 -­‐176.34 33 4/13/01 15:33:53 -­‐59.72 -­‐25.59 26 4/28/01 4:49:53 -­‐18.06 -­‐176.94 351 5/20/01 4:21:43 18.82 -­‐104.45 33 6/3/01 2:41:57 -­‐29.67 -­‐178.63 178 6/26/01 12:33:52 -­‐4.07 -­‐104.47 10 8/6/01 3:52:59 -­‐55.54 -­‐123.42 10 8/21/01 6:52:06 -­‐36.81 -­‐179.57 33 8/25/01 2:02:02 7.63 -­‐82.77 24 9/2/01 10:06:51 -­‐54.36 -­‐137.02 10 10/2/01 0:48:18 -­‐16.18 -­‐173.82 106 10/17/01 11:29:09 19.35 -­‐64.93 33 10/21/01 0:29:21 -­‐37.14 178.98 18 11/9/01 0:47:55 9.64 -­‐82.3 10 11/28/01 14:32:32 15.57 -­‐93.11 84 12/7/01 19:27:34 -­‐44.22 168.82 10 1/1/02 10:39:06 -­‐55.21 -­‐129 10 1/16/02 23:09:52 15.5 -­‐93.13 80 3/9/02 12:27:11 -­‐56.02 -­‐27.33 118 31 6.1 6.1 6.6 7.7 6.6 6 6.1 6.2 6.2 6.9 6.3 7.2 6.1 6.7 7.1 6.1 6.3 6.2 6 6.7 6.1 6.4 5.8 6 6.4 6 Table S3. Earthquakes recorded at station PEL and used in this study Date Time Lat Long Depth Mag 2/25/96 3:08:18 16.204 -­‐97.963 21 6/2/96 2:52:09 10.797 -­‐42.254 10 9/5/96 8:14:14 -­‐22.118 -­‐113.436 10 1/11/97 20:28:26 18.219 -­‐102.756 33 4/22/97 9:31:23 11.112 -­‐60.892 5 4/28/97 12:07:37 -­‐42.504 42.686 10 5/1/97 11:37:36 18.993 -­‐107.35 33 7/19/97 14:22:08 16.333 -­‐98.216 33 9/2/97 12:13:22 3.849 -­‐75.749 198 9/20/97 16:11:32 -­‐28.683 -­‐177.624 30 3/20/98 21:08:08 -­‐50.008 163.107 10 3/25/98 3:12:25 -­‐62.877 149.527 10 7/9/98 14:45:39 -­‐30.487 -­‐178.994 129 8/4/98 18:59:20 -­‐59.3 -­‐80.393 33 8/23/98 13:57:15 11.663 -­‐88.038 54 12/27/98 0:38:26 -­‐21.632 -­‐176.376 144 3/31/99 5:54:42 5.827 -­‐82.616 6/15/99 20:42:05 18.386 -­‐97.436 70 11/7/00 0:18:04 -­‐55.627 -­‐29.876 10 1/13/01 17:33:32 13.049 -­‐88.66 60 6/3/01 2:41:57 -­‐29.666 -­‐178.633 178 11/15/02 19:58:31 -­‐56.051 -­‐36.404 10 1/22/03 2:06:34 18.77 -­‐104.104 24 5/4/03 13:15:18 -­‐30.531 -­‐178.232 62 8/4/03 4:37:20 -­‐60.532 -­‐43.411 10 8/21/03 12:12:49 -­‐45.104 167.144 28 1/2/06 6:10:49 -­‐60.934 -­‐21.575 13 1/4/06 8:32:32 28.164 -­‐112.117 14 5/16/06 10:39:23 -­‐31.779 -­‐179.313 152 8/20/06 3:41:47 -­‐61.029 -­‐34.365 13 1/30/07 4:54:50 -­‐54.74 146.298 11 6/13/07 19:29:40 13.554 -­‐90.618 23 9/30/07 5:23:34 -­‐49.271 164.115 10 10/15/07 12:29:36 -­‐44.785 167.583 18 11/16/07 3:13:00 -­‐2.312 -­‐77.838 509 11/29/07 19:00:19 14.973 -­‐61.263 156 2/8/08 9:38:14 10.671 -­‐41.899 9 2/23/08 15:57:19 -­‐57.326 -­‐23.421 14 4/12/08 0:30:12 -­‐55.664 158.453 16 6/30/08 6:17:44 -­‐58.22 -­‐22.1 8 32 7.1 7 6.9 7.2 6.7 6.8 6.9 6.9 6.8 7 6.7 8.1 6.9 7.2 6.7 6.8 6.8 7 6.8 7.7 7.2 6.6 7.6 6.7 7.6 7.2 7.4 6.6 7.4 7 6.9 6.7 7.4 6.8 6.6 7.4 6.9 6.8 7.1 7 Figure S1. CCP RF stacks for the same section calculated using 3 different vp/vs ratios (1.7, 1.75 and 1.8). As was noted in the text, the features that we observe don’t change, the different vp/vs ratio simply shifts the features vertically in the section. 33 Figure S2. Ray density plots for the cross-sections shown in Figures 5, 6, and 7. Colors indicate the number of traces that are stacked per bin. Under the forearc, as many as 50 traces are being stacked in each bin, providing sufficient coverage to image the downgoing slab. 34 Figure S3. A comparison of CCP RF stacks for profile X2 for normalized and unnormalized data. They show similar results. The downgoing slab and continental Moho are observed in both sections. We present the unnormalized RFs in this manuscript. 35