Seismic Attenuation, Time Delays, and Raypath Bending of Teleseisms Beneath Uturuncu Volcano, Bolivia GEOSPHERE; V

Home , Country rock (geology)

Research Paper THEMED ISSUE: PLUTONS: Investigating the Relationship between Pluton Growth and Volcanism in the Central Andes

GEOSPHERE Seismic attenuation, time delays, and raypath bending of teleseisms beneath Uturuncu volcano, Bolivia GEOSPHERE; v. 13, no. 3 Alexandra K. Farrell, Stephen R. McNutt, and Glenn Thompson School of Geosciences, University of South Florida, 4202 E. Fowler Avenue, NES 107, Tampa, Florida 33620-5550, USA doi:10.1130/GES01354.1

11 figures; 6 tables; 1 supplemental file ABSTRACT micity even though the volcano has not erupted for 270 ka (Sparks et al., 2008). A joint geophysical experiment between scientists from the United States and CORRESPONDENCE: akfarrell@mail.usf.edu A set of 14 teleseismic earthquakes was studied to determine how wave the United Kingdom, funded by National Science Foundation and Natural Envi- propagation was affected by a presumed magma body beneath Uturuncu vol- ronmental Research Council (NERC), respectively, was conducted from 2009 to CITATION: Farrell, A.K., McNutt, S.R., and Thompson, cano, Bolivia. Teleseisms are suitable for study because they are relatively long 2014. Called PLUTONS (shortened from PLUTONNNSSSSSSSS; Probing Lazu- G., 2017, Seismic attenuation, time delays, and raypath bending of teleseisms beneath Uturuncu vol- period, contain purely P waves, and have near-vertical incidence angles. The fre and Uturuncu TOgether: NSF, NERC, NSERC, Sergeotecmin, Sernageomin, cano, Bolivia: Geosphere, v. 13, no. 3, p. 699–722, doi: number of events is small but the events have good signal-to-noise ratios and Observatorio San Calixto, Universidad Nacional de Salta, Universidad Mayor 10.1130/GES01354.1. very similar waveforms for each event so that reliable measurements could be San Andres, Universidad de PotoSi, SERNAP, Chilean Seismological Service, made of arrival times and amplitudes. Attenuation of amplitudes occurs in a Universidad de San Juan), the project included geologic (Sparks et el., 2008; Received 1 May 2016 NW-SE trend beneath the volcano, 14 by 34 km (long axis NW-SE). Calculated Perkins et al., 2016), gravity (del Potro et al., 2013), magnetotelluric (Comeau Revision received 6 December 2016 Accepted 15 February 2017 values of the quality factor Qp are an average of 12.4, with extreme values as et al., 2015), geodetic (Pritchard and Simons, 2002; Henderson and Pritchard, Final version posted with erratum 21 April 2017 low as 1.8. These calculations are based on the assumption that the highest 2013), and seismic (Jay et al., 2012; Ward et al., 2014; Kukarina et al., 2014) amplitude observed is the “true” amplitude, and all others have been attenu- components. Here we present new results based on seismic investigations of ated. The average thickness of the anomaly is 10.2 km, and the center is ~20 km teleseisms, distant earthquakes whose long-period P waves travel through the SE of the summit, within the area of surface uplift measured geodetically. Time crust at near-vertical incidence. delays of up to 0.8 s were also observed. The pattern of attenuation and rela- Determining the interaction between seismic waves and the seismic tive time delays together showed four trends: fast and not attenuated (normal low-velocity zone (and possible magma body) beneath Uturuncu volcano crust), slow and attenuated (partial melt), fast and attenuated (likely high frac- is important because such interactions affect the results of seismic imaging ture density), and slow but not attenuated (possible deep low Vp structure). methods and can be used as additional sources of information to constrain Back azimuth differences of up to 60° were observed. In nearly all cases, the location and properties of partial melt. Previous studies show that velocity azimuths were rotated into directions parallel to local rock fabric, suggesting decreases and attenuation increases as the solidus is approached and partial that shallow crustal properties affected near-surface wave propagation. Over- melting begins to occur in laboratory samples (e.g., Sato et al., 1989). We an- all results suggest partial melt as high as 10%–20% in a region of varying thick- ticipate that seismic waves will be slowed down as well as attenuated, if partial ness, low Bouguer gravity and resistivity, high Vp/Vs, persistent seismicity, melt is encountered in the crust. Attenuation at volcanic centers can be caused and overlapping a locus of recent uplift. by several factors such as presence of magma, hydrothermal systems, hetero geneous material properties, or a combination of these (e.g., Schurr et al., 2003). Therefore, if the inflation center is an area of magma injection, we ex- INTRODUCTION pect seismic waves that have passed through the inflation source to become attenuated, and therefore have diminished amplitudes compared to those that Uturuncu volcano, Bolivia, has attracted considerable scientific attention have passed through normal crust. Likewise, a crustal magma body should over the past few years. It was one of only four volcanic centers to show defor- decrease the values of the crustal P-wave quality factor, Qp, both directly in mation based on an interferometric synthetic aperture radar (InSAR) survey the body and possibly within the heated host rock surrounding the magma. of over 900 South American volcanoes by Pritchard and Simons (2002). This While seismic velocity is decreased by the presence of partial melt, especially survey was superseded by a later study showing nine deforming centers in interconnected melt, it is not particularly sensitive to temperature and there- the Central Andes (Henderson and Pritchard, 2013). Modeling showed that the fore won’t be significantly affected by rocks at near-solidus temperatures; this volcano was inflating 1–2 cm/yr, though this rate has decreased with time (Hen- is where attenuation, which is sensitive to temperature, can add information For permission to copy, contact Copyright derson and Pritchard, 2014), from a source at mid-crustal depths of 17–20 km (Lees, 2007). It has been documented that the presence of magma within the Permissions, GSA, or [email protected]. below sea level. A reconnaissance survey in 2003 showed a high rate of seis- crust can bend seismic waves, thus resulting in distorted seismic raypaths

given the locations of the hypocenters with respect to the stations (e.g., Steck by a thin film of saline fluid (Schilling et al., 2006). Gravity data analyzed by and Prothero, 1993; Yeguas et al., 2011; Galluzzo and La Rocca, 2013). This Schmitz et al. (1997) displayed a region below the magmatic arc of density can modify the results of seismic studies that rely upon detailed knowledge of 0.16–0.25 g/cm3 lower than that of the surrounding crust. They interpreted it the raypaths. In this paper, we examine velocities, attenuation, and directional as a magma body of 10–20 vol% melt that reduced P-wave velocity by 0.8– features of teleseismic waveforms to place additional constraints on magma 1.2 km/s. These geophysical studies all agree on ~10–27 vol% partial melt. body properties, location, and shape. The crust above the magma body is strongly anisotropic, with 20%–30% anisotropy in a 3-km-thick surface layer and 15%–20% anisotropy in the re- maining crust between this surface layer and the magma body; the anisotropy BACKGROUND of both crustal regions has a strike of between 300° and 330° with a tilt of 45° from vertical (slow S-wave orientation), producing significant azimuthal Uturuncu is located in SW Bolivia at latitude 22.27°S and longitude 67.18°W. variations in the radial receiver function components (Leidig and Zandt, 2003; The volcano is also at the SW part of two major provinces, one based on geol Zandt et al., 2003). Anisotropy likely results from alignment of a system of ogy (the Altiplano-Puna volcanic complex [APVC]) and the other based on fluid-filled cracks oriented ENE/WSW and dipping to the north from 45° to 80° the presence of a seismic feature at depth (the Altiplano-Puna magma body from horizontal (Leidig and Zandt, 2003), or a crustal strain partitioning into [APMB]). These are subfeatures of the much larger Altiplano-Puna plateau. subhorizontal zones of high and low strain (Schilling et al., 2006). The pres- The Bolivian Altiplano is part of the 1800-km-long, 350–400-km-wide Alti ence of a large crustal magma body in the Andes, an area of strong compres- plano-Puna plateau, enigmatically formed in the absence of continental colli- sive tectonics, suggests that the APVC may be experiencing a transition from sion (e.g., Allmendinger and Gubbels, 1996). Uplift began ca. 25 Ma, as the vertical thickening to horizontal extension along the arc (Riller et al., 2001). convergence rate between the Nazca and South American plates increased and Vertically, this magma body also forms at the boundary between upper-crustal subduction shallowed (Allmendinger et al., 1997). Presently, the subduction imbrication and lower-crustal thickening (Yuan et al., 2000). angle is 30° beneath the Altiplano, which has a lithospheric thickness, attrib- There are three existing seismic models for the geometry of the APMB. utable to crustal thickening, of up to 150 km (Allmendinger et al., 1997). The The first is that of Zandt et al. (2003), based on receiver functions, in which average Vp/Vs ratio of the southern Altiplano is 1.80, suggesting that the seismic the magma body is at 19–20 km depth below the surface, 1 km thick, and is velocities have been affected by high temperatures and regions of partial melt- a compositionally uniform lens (see Fig. 1). Here, the APMB underlies an 2 ing (Yuan et al., 2000). These mainly lower the Vs (e.g., Nakajima et al., 2001). ~60,000 km area of 3° in longitude and 2° in latitude (Zandt et al., 2003). The The APVC and APMB are two distinct features of the Central Andes that second model is the joint ambient noise tomography and receiver function are affected by stresses in addition to subduction. The APVC is a major silicic result of Ward et al. (2014), in which the magma body is ~150 km across and volcanic province, located between 21° and 24°S (333 km) and 66° and 69°W centered beneath the center of surface InSAR-detected deformation (Fig. 1). (307 km), and resulting from an ignimbrite “flare-up” beginning in the late The thickness of the central, strongest part of the anomaly is ~11 km, though Miocene. It is deemed to be active because of Late Pleistocene and younger there is an anomaly from 4 to 25 km below sea level. In this model, the APMB volcanic activity and active low-temperature geothermal fields (de Silva, 1989; is shallowest (4 km depth to surface) below Uturuncu, and the volume of the de Silva et al., 1994). It is the youngest but largest of several such ignimbrite magma body is given as ~500,000 km3. In this model, the APMB contains ~4%– fields in the Central Andes volcanic zone, covering an area of ~60,000 km2 lo- 25% partial melt (Ward et al., 2014), a value that is consistent with resistivity es- cated at the transition from the ~4 km high Altiplano south through the ~5-km- timates of at least 20% partial melt (Comeau et al., 2015). The third is a model high Puna (de Silva, 1989; Whitman et al., 1996). of local thickening in two rectangular blocks of ~40 km deep by 20 km wide The APMB was identified as a ~20-km-deep, 1-km-thick zone of low seismic (shallower block) and ~30 km deep by 10 km wide (deeper block) beneath the

velocity, for which Vs < 1.0 km/s (Zandt et al., 2003). It mostly underlies the volcano, as recovered by Vp/Vs tomography from Kukarina et al. (2014; shown <7 Ma ignimbrite complexes of the APVC, among which Uturuncu is located, schematically in Fig. 1). rather than the Quaternary arc volcanoes to the west. The APMB is coincident Local thickening of the APMB beneath Uturuncu volcano, as observed by with an ultra-low velocity zone (ALVZ), a region characterized as causing about Comeau et al. (2015), may be the result of a diapir (Fialko and Pearse, 2012). a 10%–20% reduction in seismic-wave velocity over a thickness of 10–20 km However, magnetotelluric data suggest that the magma that sources this (Yuan et al., 2000). Schilling et al. (1997) discovered a region of high con- APMB upwelling exhibits a greater resistivity than that of the APMB proper. ductivity (1 S/m) from depths of ~20 km to at least 60 km below the Western This may reflect a different composition, differing melt connectivity, or a lower Cordillera, which they modeled as the result of a body of 14–27 vol% inter- melt fraction in this source magma (Comeau et al., 2015). In addition, above connected partial melt. Using temperature modeling, heat-flow densities, and 10 km below sea level, magnetotelluric studies show that crustal magma eruptive history, Schilling and Partzsch (2001) determined that this melt most distribution is not symmetric and occurs in discrete bodies (Comeau et al., likely has a granite-like composition and a crustal origin. It is likely overlain 2015; Comeau et al., 2016). In the magnetotelluric study, the APMB is located

A Uturuncu volcano A′

0 Ward et al., 2014, model (4–25 km)

Zandt et al., 2003, model (19/20–21 –10 km from surface, 1-km-thick layer)

–20 Figure 1. True-scale east to west cross section from A to A′ (see Fig. 2) of different models of the Altiplano-Puna magma Comeau et al., 2016, model for body (APMB). Light-gray model bounded –30 APMB (10–15 km depth with by long dashed lines is that of Ward et al. indeterminate thickness, with a (2014); the dark-gray model bound by shallow volcanic magma chamber short dashed lines is that of Zandt et al. –40 which is not shown). This is the (2003); the unlined dark-gray model is main C2 anomaly—there are that of the C2 conductor of Comeau et al. Depth (km) conductive areas laterally as well. (2016), and the vertically extensive gray shape beneath the volcano is the model of –50 Kukarina et al. (2014). Depth to the Wadati- Kukarina et al., 2014, model (vertically Benioff Zone from Cahill and Isacks (1992). extensive, narrow Vp/Vs anomaly) Uturuncu volcano is the labeled topographic feature at center. –60

–70 Base of crust (75 km) Wadati-Benioff Zone (150 km) –80 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Distance along profile (°)

at 16–20 km below the surface, with a resistivity of <3 W m centered ~3 km of the Andean volcanic arc (Allmendinger et al., 1997). Local seismicity has W of the summit and extending off eastward. There is a vertically elongated been recorded since seismic networks were installed in 2009 (details below). low-resistivity body directly beneath the volcano that can be attributed to the The volcano was last active from 890 to 270 ka (Sparks et al., 2008), erupting magma chamber discussed in Muir et al. (2014). The resistivity of this body magmas of dacitic and andesitic compositions. Current local, low-level ther- suggests saline aqueous fluids in addition to 15% melt, as even pure melt is mal activity is characterized by two active fumarole fields near the volcano’s insufficiently conductive (Comeau et al., 2015; Comeau et al., 2016). Highlight- summit (Sparks et al., 2008). Despite its dormancy, recent InSAR studies show ing narrow, vertically elongated structures in the upper crust above the APMB, a maximum inflation rate of 1.5–2 cm/yr over an area of 70 km width (Pritchard del Potro et al. (2013) modeled the APMB as a body of non-uniform thickness and Simons, 2002) and which is surrounded by subsidence to create a som- with diapirs extending toward the surface from the top of the magma body brero-shaped deformation pattern (Fialko and Pearse, 2012; Henderson and (Fig. 1). The body thickens beneath these diapirs and thins away from them. Pritchard, 2013; Hickey et al., 2013). This inflation was initially modeled as The model they favor shows a 25% melt fraction, though they agree that the a point source with a depth of 15–17 km below sea level located 3 km to the APMB more likely contains varying melt fractions. southwest of Uturuncu’s summit (Pritchard and Simons, 2002). However, current interpretations are that the deformation signature, with the associated Uturuncu ring of subsidence, is the result of buoyant magma at a depth of 19 km below the surface, bulging upward at the center of deformation, displacing hot and Uturuncu volcano (Fig. 2) is a 6008-m-tall stratovolcano located in the ductile crustal rocks which then flow aside and downward (Fialko and Pearse, Central Volcanic Zone of the Andes. Uturuncu is located upon ~70-km-thick 2012). In this model, magma is pulled laterally into the diaper at depth. Current crust ~150 km above the top of the subducting slab, 50 km east of the axis studies show a source volume of decreased density with respect to the crust,

67°30′0″0WW 67°15′0″W 67°0′0″W 66°45′0″W Elevation PLLB 6008 PLCL

5559 PLKN 5119 PLHS 22°0′0″S PL03 22°0′0″S 4679 B PLJR PLTP 4239 PLTM meters N PLVB 3799 PLSM

PLMN PLDK PLLC PLQU

PLCM PLMK 22°15′0″S PLRR 22°15′0″S A A′ PLRV PLBR PLLA PLSS PLLO PLWB PL07 PLSQ PLLL PLSE PLCO PLAR 22°30′0″S 22°30′0″S PLMD

PLTT B′ PLAN PLSP Kilometers 084 16 24 32

67°30′0″W 67°15′0″W 67°0′0″W 66°45′0″W

Figure 2. Location of seismic stations (black circles) in the Probing Lazufre and Uturuncu TOgether: NS F, NERC, NSERC, Sergeotecmin, Sernageomin, Observatorio San Calixto, Universidad Nacional de Salta, Universidad Mayor San Andres, Universidad de PotoSi, SERNAP, Chilean Seismological Service, Universidad de San Juan (PLUTONS) seismic network around Uturuncu volcano (white triangle), on a background of colored elevation. Stations with two names indicate that the station was moved >50 m and therefore renamed at some point in the deployment. Location of seismic network is shown in red box (arrow) in the index map. Line from A to A′ shows the location of the cross section in Figure 1 and line from B to B′ shows the location of the cross section in Figure 5.

a high Vp/Vs ratio of >1.9, and a depth to the top of the layer of ~6 km below body (Comeau et al., 2016). Ward et al. (2014) used the results of a combined the surface (e.g., del Potro et al., 2013; Kukarina et al., 2014; McFarlin et al., receiver function and surface wave dispersion study (Ward et al., 2013) to de- 2014). The greater depth (16–20 km below surface) of the APMB’s magneto- termine the location and size of the APMB. The Ward et al. (2014) study is telluric signature is attributable to possible layering of different compositions regional scale, smooths the anomalies over a broad lateral area while preserv- (i.e., magmas of differing resistivities) in the magma body, which would push ing vertical resolution, and focuses on results derived from velocities. Here the magnetotelluric depths to greater depths than the actual top of the magma we seek more detailed information on the smaller region around Uturuncu,

as well as adding to the understanding of how the anomaly affects seismic TABLE 1. SEISMIC STATION INFORMATION amplitudes via attenuation. Elevation Distance* Azimuth§ Uturuncu volcano exhibits a remarkable lack of variation in composition Station nameLatitudeLongitude (km) (km) (°) throughout its eruptive history (Sparks et al., 2008; Muir et al., 2014). Although PL03 –22.0156 –66.9451 4.63 37.2 41 the volcano is compositionally dacitic, it shows no evidence of having erupted PL07 –22.3938 –67.0205 4.64 21.4 130 explosively (Sparks et al., 2008). This follows reasoning by de Silva (1989) PLAN –22.6445 –67.3866 4.70 46.8 207 that intrusion rate into the crust has a stronger effect on eruptive style than PLAR –22.4873 –66.9778 4.63 31.9 139 PLBR –22.3055 –67.2356 4.80 7.0235 magma composition or volatile content. The composition of erupted products PLCL –21.9065 –67.0294 4.35 43.3 21 at Uturuncu is likely the result of both the formation of dacite by fractional PLCM –22.2419 –67.2054 5.15 4.2320 crystallization of andesitic parent lava within the APMB and magma mixing in PLCO –22.4615 –67.4372 4.69 34.0 231 a shallow storage zone 5–7 km below Uturuncu’s summit (Muir et al., 2014). PLDK –22.1551 –66.9784 4.55 24.4 58 Their study informs on both the depth at which magma is stored preerup- PLHS –21.9566 –67.2594 4.53 35.8 347 tively in the Andean crust beneath Uturuncu volcano and the composition PLJR –22.0475 –67.3487 4.24 30.3 325 PLKN of the melt, which impacts factors such as the density and bulk modulus of –21.9576 –67.2604 4.54 35.7 347 PLLA –22.3126 –67.3796 4.24 21.1 257 magma being encountered by seismic waves. Geothermometry indicates that PLLB –21.8955 –67.3796 4.09 46.5 334 temperature varies temporally and spatially within the magma body feeding PLLC –22.1415 –67.5879 4.75 44.4 289 Uturuncu and that the magma may experience large temperature fluctuations PLLL –22.4205 –67.1500 4.75 17.1 170 prior to eruption (Muir et al., 2014), suggesting that magma storage conditions PLLO –22.3337 –67.0793 4.58 12.6 124 are non-uniform. PLMD –22.5490 –67.1912 4.63 31.1 182 Seismicity at Uturuncu consists of shallow, near sea level (~5 km beneath PLMK –22.2515 –67.0769 4.48 10.9 79 PLMN –22.1607 –67.1207 4.51 13.7 27 the volcano’s summit), almost exclusively volcano-tectonic events; however, PLQU –22.1906 –67.3389 4.15 18.6 298 some events with low frequencies appear to be deeper (>20 km) events whose PLRR –22.2611 –66.8820 4.48 30.7 88 higher-frequency components have been attenuated (Jay et al., 2012; Kukarina PLRV –22.2843 –67.5933 4.87 42.6 268 et al., 2014). Seismic swarms were observed a few times a month, consisting PLSE –22.4466 –67.2902 4.68 22.7 210 of 5–60 events occurring over a time span of minutes to hours. One swarm PLSM –22.1130 –67.2844 4.22 20.5 328 appears to have been triggered by the 2010 Mw = 8.8 Maule earthquake. This PLSP –22.6698 –67.1208 4.73 44.9 172 PLSQ suggests that the hydrothermal system at the volcano is metastable, being –22.4197 –67.6381 4.81 50.0 250 PLSS –22.3168 –67.1715 4.69 5.4170 affected by stresses caused by magma accumulation at ~20 km depth within PLTM –22.0634 –67.1671 4.70 23.0 3 the mid-crust (Jay et al., 2012). The calculated b-value is low, within the range PLTP –22.0413 –66.7865 4.39 47.9 58 for tectonic earthquakes at 0.64 ± 0.04 (Jay et al., 2012). Analysis of regional PLTT –22.5937 –67.2790 4.64 37.4 196 slab earthquakes showed variable amounts of S-wave attenuation (M. West, PLVB –22.0632 –67.1676 4.71 23.1 3 2016, written commun.) suggesting that waves from different earthquakes ex- PLWB –22.3537 –66.8347 4.49 36.7 105 perience varying amounts of attenuation in their respective paths (Jay et al., Note: Station latitudes are in °S; longitudes in °W, and elevations in km above sea 2012). Jay et al. (2012) also used heat-flow density to determine that the depth level. *Distance is between the station and the summit of the volcano. to the brittle-ductile transition zone beneath Uturuncu is at approximately sea §Azimuth is from the volcano’s summit to the station and is measured in degrees level (6 km below the summit), though it could be as deep as 8.5 km below clockwise from north. sea level. Because Jay et al. (2012) favor the shallower depth, earthquake activity apparently persists into the inferred brittle-ductile transition zone. ber 2012. Dates of operation and numbers of stations are given in Figure 3. Each station consisted of a Guralp CMG3T three-component broadband seis- DATA AND METHODS mometer, a separate GPS clock, and a RefTek RT 130 Datalogger. Data were collected at sampling rates of 100 samples/s and 1 sample/s (same data but The PLUTONS network (Fig. 2 and Table 1) consisted of 28 seismom- with different sampling frequencies) and were recovered during service runs eters deployed between April 2010 and October 2012. Some stations of the every six months. PLUTONS array were co-located with previous ANDIVOLC stations (Jay et al., In the current work, we studied 14 teleseismic events from several subduc- 2012). The first half of the network was installed in April 2010 and removed in tion zones at different azimuths with respect to the network in Bolivia (Fig. 4 and March 2012; the second half was installed in April 2011 and removed in Octo- Table 2). Deep events were chosen to ensure that the seismic waves have only

lyzed both PKIKP and PKP phases to determine if a differing angle of incidence and travel path gave different results. To ensure that the same P-wave phase of each teleseismic earthquake was analyzed for every station, we matched the phase by aligning waveforms using cross correlations of a 10–20 s win- dow of data filtered from 0.375 Hz to 1.5 Hz across the network and common waveform characteristics, such as wave shape and frequency. Traces in which the phase could not be identified were discarded manually. Correlation coeffi cients ranged mostly from 0.72 to 0.949 (Table 3 and Table S11) with 1 being perfect correlation. One station, PLCO, had systematically low values. We then compared our calculated amplitudes across the network to determine if, for a N given source, there is a corresponding area of diminished amplitude where the waves have passed through an attenuating body. As a control, and because few studies like this are known to us, we also performed this method using a

1Supplemental Materials. A map showing the results teleseismic event from Bolivia on Transportable Array (TA) stations in Florida of testing our method of determining amplitudes in a (Fig. S1 [see footnote 1]) to show how larger-scale variations could be resolved different location, additional seismograms, and addi- with this technique. Details are given in the discussion below and in the Sup- tional amplitude vs. time graphs. Please visit http:// doi.org/10.1130/GES01354.S1 or the full-text article plemental Materials (see footnote 1). We identified seismic phases through on www.gsapubs.org to view the Supplemental TauP modeling (University of South Carolina [USC]; www.seis.sc .edu /taup/) of Materials. arrival times. The phases are shown in Table 2. Examples of the seismograms for one event are shown in Figure 6; seismograms from all other events are shown in the Supplemental Materials (see footnote 1). To calculate the P-wave quality factor (Qp) values within the upper 30 km of Figure 3. Seismic station coverage for the duration of the Probing Lazufre and Uturuncu TOgether: NS F, NERC, NSERC, Sergeotecmin, Sernageomin, Observatorio San Calixto, Universidad Nacional de Salta, each ray’s travel path, we used the formula: Universidad Mayor San Andres, Universidad de PotoSi, SERNAP, Chilean Seismological Service, Univer- sidad de San Juan (PLUTONS) deployment at Uturuncu. Squares show deployment dates, diamonds −π∗()Rf /V denote recovery dates. Black lines represent coverage while gray lines show time periods of data drop- Qp = , (1) ln()AA/ 0 out. Note that station PLHS is a relocated PLKN, and station PLVB is a relocated PLTM.

where R is distance traveled in the attenuating medium (20 km for this study), traveled once through the crust and asthenosphere in the vicinity of Uturuncu f is frequency of the wave, V is velocity (average of local velocity model of 5.15

and, thus, were not affected by the crust or asthenosphere near the earthquake km/s), A is the amplitude at each station, and A0 is the maximum amplitude sources. Slab events of magnitude 5 and greater are visible in the Uturuncu measured for the earthquake. We assume the wavefronts are planar at teleseismic data with good signal-to-noise ratios. The teleseismic waveforms gen- seismic distances; hence, we ignore geometric spreading. The calculations are erally have periods of several seconds; thus, the wavelengths are of the order of based on the assumption that the highest amplitude observed is the “true” 10 km and not sensitive to fine structure. The incidence angles are steep (Fig. 5); amplitude, and all others have been attenuated. Thus, this formula results in

so the P waves mainly sampled vertical structures. The teleseisms show sys- an apparent infinite Qp value for the station with largest amplitude A( = A0); so tematic changes between waveform peak-to-peak amplitudes and time resid- that value was not included in subsequent analyses. To determine if crustal Qp uals as a function of different earthquake back azimuths. We used events from values are frequency dependent, we analyzed the amplitudes of each wave- the European subduction zones (ESZ; one event, the Calabria subduction zone), form at different filters in octave steps from 0.375 Hz to 1.5 Hz (0.375–0.75 Hz >80 km depth for the Kermadec-Tonga subduction zones (KTSZ; four events), and 0.75–1.5 Hz), and then over the full range of 0.375–1.5 Hz. Then, we used

>400 km depth for the Japan subduction zone (JSZ; four events), and >100 km the relationship between Qp, ln(A/A0), and f (the center frequency of the band- for the South Sandwich Islands subduction zone (SSSZ; five events). Results pass filter) to determine how a change inf affects Qp. This evaluation was from the ESZ were difficult to constrain given relatively low signal-to-noise done using the waveform suite for MATLAB (Reyes and West, 2011). ratio and therefore only included in the discussion of amplitudes. The P waves travel through the region of interest at depth (presumed to We used a combination of MATLAB and Antelope (Boulder Real Time be a magma body) but also travel through the local shallow crust. H. McFarlin Technologies [BRTT]; www.brtt.com) to calculate the P-wave peak-to-peak and (2014, personal commun.) followed the methodology of Frankel (1982) and zero-to-peak amplitudes of the vertical component at each station for each used local shallow earthquake data to determine the average Qp value for the earthquake. For an event from Japan (JSZ2a and JSZ2b; Table 2), we ana- local crust and obtained a range of values from 60 to 770, with most values (16

0 165

330 144 30

127.5

98 90 300 60

52.5

Figure 4. Map, centered on Uturuncu volcano, showing the events used in this study (Table 2; red stars) and their 15 event-station paths (red lines). Black radial lines are gridded every 30°, and black con- centric circles gridded every 37.5°, starting 270 90 at 15° from the volcano. Blue circle shows the maximum expected epicentral dis- 15 tance for which to find PcP arrivals, and the green line the epicentral distance at

which Pdiff is no longer the first arrival, derived from the IASP91 velocity model.

52.5

240 120 90 98

127.5

210 144 150

165 180

out of 26 total measurements) falling within the range of 130–240. There were the deviation in relative arrival times from a plane wave arriving at the geo- no systematic trends observed in these data. Hence the shallow crust causes graphically closest station earliest, we first created a plane using the coordi- a modest reduction in all amplitudes, but we infer that the main deviations are nates of the station closest to the earthquake epicenter as well as the known caused by the presence of a magma body at depth. back azimuth and angle of incidence. Then, we calculated the shortest distance We also determined relative arrival times for each event, assuming straight between the plane and each station using the point-plane distance calculated rays through a 2-D velocity model. To determine delay times, here defined as from the Hessian Normal Form of the plane. Using a 2-D P-wave velocity model

TABLE 2. CATALOG OF SEISMIC EVENTS STUDIED IN THIS PAPER Time Arrival time* Depth Latitude Longitude Period AOI Azimuth Name Date (UTC) (UTC)MM type (km) (°) (°) ∆ (s) Phase (°) (°) ESZ1 5/19/11 14:50:54 15:02:10 4.7 Mw 29239.21 14.9698.10.9 Pdiff 13 51 KTSZ1 4/18/11 13:03:02 13:16:20 6.6 Mw 86 –34.34 179.87 94.8 1.4 PcP 13 230 KTSZ2 7/29/11 7:42:22 7:55:20 6.7 Mw 532–23.78 179.76 100.32.5 Pdiff 13 239 KTSZ3 9/15/11 19:31:04 19:43:50 7.3 Mw 645–21.61 –179.53100.82.6 Pdiff 13 241 KTSZ4 1/24/12 0:52:05 1:04:55 6.3 Mw 580–24.98 178.52 100.72.6 Pdiff 13 237 JSZ1 11/30/10 3:24:41 3:43:42 6.8 Mw 48628.36 139.15 155.51.3 PKIKP 4290 JSZ2a 1/12/11 21:32:55 21:51:50 6.4 Mw 52726.98 139.87 155.01.5 PKIKP 4286 JSZ2b 1/12/11 21:32:55 21:52:17 6.4 Mw 52726.98 139.87 155.01.05 PKP 13 286 JSZ3 5/10/11 15:26:05 15:44:55 5.4 Mb 54443.29 130.94 154.21.3 PKIKP 4329 JSZ4 10/4/11 1:37:29 1:56:30 5.6 Mw 45526.77 140.43 154.51.3 PKIKP 4286 SSSZ1 1/20/11 3:44:26 3:52:45 5.2 Mb 129–59.94 –27.48 46.8 1.15 P 24 154 SSSZ2 1/23/11 22:53:02 23:01:04 5.3 Mb 124–56.47 –26.96 45.1 1.3 P 24 150 SSSZ3 6/19/11 8:37:45 8:45:45 5.2 Mw 119–56.05 –27.42 44.6 1.3 P 24 149 SSSZ4 8/21/11 12:38:54 12:46:52 5.6 Mw 130–56.43 –27.49 44.8 1.3 P 24 150 SSSZ5 8/26/11 7:41:23 7:49:23 5.2 Mb 127–56.21 –27.18 44.8 1.4 P 24 150 *Denotes arrival time at the Probing Lazufre and Uturuncu TOgether (PLUTONS) seismic network in Bolivia. Abbreviations: ESZ—European subduction zone; KTSZ—Kermadec-Tonga subduction zone; JSZ—Japan subduction zones; SSSZ—South Sandwich subduction zone. M—magnitude; ∆—distance from event to the volcano’s summit in degrees; AOI—angle of incidence, from vertical; azimuth is measured from the volcano’s summit to the earthquake’s epicenter, in degrees clockwise from north.

NW-SE earthquakes Uturuncu PLBR PLMK PL07 PLAR PLSP 5.4 PLJR PLSM PLMN PLSS PLLL 3.4

1.4

–0.6

–2.6

–4.6

–6.6 Figure 5. Incident rays for two teleseisms –8.6 (JSZ from NW and SSSZ from SE) on a cross section from B to B′ (Fig. 2) show- –10.6 ing steep incidence angles. Red arrows –12.6 show attenuated signals, black arrows

Elevation (km) unattenuated. Wavefronts of the inferred –14.6 plane waves are perpendicular to the rays –16.6 shown. Horizontal black lines are for reference and are 5 km apart. –18.6

–20.6

–22.6

–24.6

–26.6

–28.6

–30.6 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 B B′ Distance along profile (°)

TABLE 3. SELECTED CROSS-CORRELATION COEFFICIENTS FOR EVENT KTSZ2 Stations PLAN PLCO PLLC PLTP PLSE PLSP PLQU PLBR PLLL PLJR PLSS PLSM PLAR PL07 PLMK PLAN 1.00 0.71 0.80 0.723 0.83 0.85 0.74 0.77 0.71 0.76 0.60 0.85 0.73 0.75 0.70 PLCO 0.71 1.00 0.57 0.389 0.60 0.56 0.41 0.58 0.66 0.51 0.51 0.58 0.46 0.40 0.32 PLLC 0.80 0.57 1.00 0.799 0.91 0.75 0.78 0.81 0.68 0.76 0.70 0.84 0.76 0.74 0.78 PLLA 0.86 0.61 0.89 0.790 0.95 0.86 0.87 0.78 0.74 0.79 0.67 0.90 0.73 0.78 0.80 PLSE 0.83 0.60 0.91 0.830 1.00 0.86 0.85 0.82 0.81 0.80 0.77 0.88 0.79 0.78 0.84 PLSP 0.85 0.56 0.75 0.713 0.86 1.00 0.82 0.76 0.72 0.66 0.68 0.82 0.77 0.75 0.77 PLQU 0.74 0.41 0.78 0.840 0.85 0.82 1.00 0.82 0.62 0.72 0.73 0.82 0.74 0.80 0.81 PLBR 0.77 0.58 0.81 0.786 0.82 0.76 0.82 1.00 0.76 0.68 0.81 0.73 0.81 0.81 0.81 PLLL 0.71 0.66 0.68 0.734 0.81 0.72 0.62 0.76 1.00 0.77 0.79 0.74 0.65 0.67 0.70 PLJR 0.76 0.51 0.76 0.862 0.80 0.66 0.72 0.68 0.77 1.00 0.64 0.89 0.61 0.84 0.72 PLSS 0.60 0.51 0.70 0.838 0.77 0.68 0.73 0.81 0.79 0.64 1.00 0.58 0.72 0.65 0.79 PLSM 0.85 0.58 0.84 0.786 0.88 0.82 0.82 0.73 0.74 0.89 0.58 1.00 0.65 0.86 0.73 PLAR 0.73 0.46 0.76 0.728 0.79 0.77 0.74 0.81 0.65 0.61 0.72 0.65 1.00 0.70 0.80 PL07 0.75 0.40 0.74 0.822 0.78 0.75 0.80 0.81 0.67 0.84 0.65 0.86 0.70 1.00 0.83 PLMK 0.70 0.32 0.78 0.898 0.84 0.77 0.81 0.81 0.70 0.72 0.79 0.73 0.80 0.83 1.00 PLDK 0.79 0.48 0.86 0.923 0.90 0.76 0.82 0.77 0.75 0.91 0.70 0.88 0.76 0.86 0.89 PLWB 0.85 0.57 0.86 0.801 0.90 0.86 0.79 0.83 0.76 0.79 0.68 0.85 0.88 0.83 0.82 PLRR 0.68 0.29 0.77 0.949 0.79 0.72 0.82 0.79 0.65 0.75 0.83 0.69 0.78 0.77 0.92 PLCL 0.73 0.48 0.82 0.876 0.82 0.68 0.84 0.81 0.68 0.81 0.68 0.80 0.75 0.81 0.81 PL03 0.71 0.36 0.82 0.930 0.80 0.71 0.83 0.80 0.64 0.77 0.77 0.74 0.76 0.77 0.86 PLTP 0.72 0.39 0.80 1.000 0.83 0.71 0.84 0.79 0.73 0.86 0.84 0.79 0.73 0.82 0.90 Note: These values are for waveforms filtered from 0.375 Hz to 1.5 Hz. Table S1 (see text footnote 1) shows all cross-correlation coefficients for this event.

modified from Ward et al. (2014), we determined the time expected for this 100% melt. To determine the velocity for each percent partial melt, we com- plane wave to reach each station as our expected times. We subtracted these puted the Hashin-Shtrikman-Wapole bounds (Hashin and Shtrikman, 1963; expected times from the actual arrival times. This procedure returned delay Mavko et al., 2009) to calculate the minimum and maximum rigidity and bulk values that are positive for a late arrival, negative for an early arrival, and al- moduli bounds as a function of percent melt. These values, combined with the ways zero at the station closest to the epicenter. The absolute value of the density of the melt body, were used to calculate P-wave velocities for each inte- lowest negative value was then added to each delay time to determine the ger melt percentage. Using this calculation, we obtained velocity values from delay with respect to the station at which the earthquake arrived the earliest 5.6 km/s for 0% melt and 4.5 km/s for 100% melt for a dry melt. Incorporating compared to the expected time. We interpret this station to be the one least af- 8 wt% water into the magma, we obtained velocities between 5.6 km/s and 2.9 fected by the magma body, which has been imaged to underlie the entire array. km/s. We used the lower bounds because crustal melts are likely to be inter- From these delay values, we determined percent partial melt using the ve- connected (Schilling and Partzsch, 2001; Comeau et al., 2016). This was done locities of P waves in melt and in solid rock. We used a density of 2300 kg/m3, for a mixture of melt and solid rock, then for this combination but with of water the value determined as the density of the region of partial melt by del Potro constituting 8 wt% of the melt fraction (Sparks et al., 2008). Using rigidities of et al. (2013), and the partial melt values using this density are reported in this 0 GPa for the melt and water fractions results in a lower rigidity bound of 0 GPa paper. We also used the crustal density value of 2800 kg/m3 (Perkins et al., and, therefore, when using this to calculate velocities, causes the velocity to 2016), though this value only decreased the partial melt percentage necessary sharply decrease with the inclusion of any the fluid. This is our reason for using by 1%–3% for a thickness of 20 km (increasing slightly with increasing percent nonzero rigidities for the water and melt. Then, the thickness of the magma melt). We employed typical crustal values of rigidity (26 GPa) and bulk modulus body was given (we tested values of 1, 10, and 20 km), and we calculated the (45 GPa), which give velocities that agree with Uturuncu’s velocity model. For time taken to travel this distance with a velocity equivalent to that in solid rock the melt, we used a bulk modulus of 45 GPa but a rigidity of 0.5 GPa, and for plus the time delay to determine our control values. Then, we used the veloci- the model including water, we calculated water as 8% of the melt fraction, with ties through melt to calculate the travel time expected for the above distance at a bulk modulus of 2.2 GPa and a rigidity of 0.1 GPa. These values are for pure each percent partial melt. For each station, the travel time value that matched to melt and pure water in a silicate melt, allowing us to mix the moduli for melt within a tolerance threshold (half of the difference between travel times for ad- (with water) and calculate velocities for a range from a purely solid crust to jacent percent partial melt values) of the control value for that station gave the

JSZ1

200 0 HH Z PLSQ −200 200 0 HH Z PLQU −200 200 0 HH Z PLLA −200 200 0 HH Z PLSM −200 200 0 HH Z PLTM −200 200 0 HH Z PLSE Figure 6. Example of teleseismic wave- −200 forms: Vertical components of event JSZ1, 200 ordered by increasing distance, with the peak used for amplitude measurements 0 HH Z

PLCM denoted by vertical black solid lines. The −200 vertical scale is equal for all stations, and is in units of nm/s. Note the very similar 200 waveforms at all stations. 0 HH Z PLSS −200 200 0 PLLL HH Z −200 200 0 HH Z PLMK −200 200 0 HH Z PL03 −200 200 0 HH Z PLAR −200 200 0 HH Z PLRR −200 03:43:35 03:43:36 03:43:37 03:43:38 03:43:39 03:43:40 03:43:41 03:43:42 03:43:43 03:43:44 03:43:45 Time

average percent partial melt value of the magma body encountered along that where N is the amplitude on the north-south component, and E is the ampli- raypath. The results are given in Table 4. Additionally, we were able to solve for tude on the east-west component. Then, we compared these measured azi- magma body thickness given percent partial melt values by using the equiva- muths to the expected back azimuth given the earthquake hypocentral location lent partial melt velocities. As a test, we used depth values from sea level to the relative to the volcano. upper surface of the magma body calculated by receiver functions (McFarlin et al., 2014) and used the bottom depth of the Ward et al. (2014) model (25 km below sea level) to determine a magma body thickness and calculate percent RESULTS partial melt values for this magma body geometry (Table 4). Initial angles of incidence were calculated using the locations of the earth- We discovered a persistent region of reduced amplitudes and delayed ar- quake hypocenters with relation to Uturuncu volcano in the center of the seis- rival times centered near the station PLAR, SE of the volcano (Fig. 7) when ana- mic array. However, when calculating delay times for events from the KTSZ and lyzing teleseisms from the JSZ and SSSZ, from the NW and SE of the volcano, SSSZ, we discovered a dependence on lag time with distance, suggesting that respectively. For these events, the attenuating zone is a 14 by 34 km (preferred the calculated angle of incidence differs from the actual incoming raypath. Using values based on station distribution) elliptical region southeast of the summit the correlation matrix methodology of Jurkevics (1988), this issue was resolved, centered between stations PLAR and PLSS (Fig. 7). This is 31° clockwise and 6°

and therefore these calculated angles of incidence were favored in the analysis. counterclockwise off of the median back azimuth for JSZ back-azimuth sets AJSZ

Finally, we determined the back azimuth—the angle clockwise from hori- and BJSZ, respectively, and 5° counterclockwise and 27° clockwise off of those

zontal from the station to the event—for each of the above teleseismic events. for the SSSZ sets ASSSZ and BSSSZ (Fig. 7). In addition, there is a region of low We calculated the peak-to-peak amplitudes of each component for all stations. amplitude continuing to the northwest of the volcano in the events coming from Using these values, we calculated each back azimuth as arctan (N/E) for events the SSSZ. This is ~20 km by 23 km, centered on a line between station PLJR and from the ESZ and KTSZ and arctan (E/N) for events from the JSZ and SSSZ, Uturuncu’s peak, and is 3° clockwise off of the median back azimuth for SSSZ

TABE 4. ESTIMATED PARTIA MET AND EOCIT REDUCTION AUES FOR SEECT EENTS 20 kmReceiver function thickness 10 km 1 km elocity elocity elocity elocity reduction No water water reduction No water water reduction No water water reduction No water water Station ppm ppm ppm ppm ppm ppm ppm ppm

KTS2 P07 6.8 33 8.7 4313.5 10 660.0>100# >100 PAR 16.8 24 9 18.3 45 10 28.6 >100 27 79.1 >100 >100 PBR 8.7 43 9.2 5315.0 15 763.2>100>100 P 12.6 96 14.9 14 722.6>1001673.8>100>100 PSS 8.7 43 10.7 6416.2 22 865.5>100>100 PTP 11.1 76 16.3 22 820.6>1001371.0>100>100 S3 PAR 14.7 13 7 16.0 19 824.3>1002176.3>100>100 PBR 15.0 16 7 17.6 32 10 25.4 >100 23 77.3 >100 >100 P 12.6 95 14.5 12 721.3>1001573.0>100>100 SSS4 PAR 7.6 43 8.7 4314.5 15 762.8>100>100 PC 8.9 53 10.7 6417.1 27 966.3>100>100 P 15.2 19 8 18.1 42 10 26.6 >100 25 78.3 >100 >100 PSS 7.0 32 8.7 4313.3 11 660.6>100>100 Note: Table shows that the higher the percent partial melt, the more strongly impacted it is by water content. Only integer partial melt values were considered. This column and those with 10 km and 1 km show the percent partial melt for an anomaly thickness of 20 km, 10 km, and 1 km, respectively. This column shows percent partial melt values epected for an anomaly with thickness calculated using the base of the ard et al. 2014 model 25 km below sea level and an upper surface at depths given by receiver function data McFarlin et al., 2014. This value eceeded 100 partial melt. It is impossible, given the parameters used and the delay time, for the anomaly to be limited to this thickness.

67.5°W 67°W 66.5°W 67.5°W 67°W 66.5°W AB B JSZ N N PLCL PLCL

22°S 22°S 22°S 22°S A PL03 PL03 JSZ PLJR PLTP PLJR PLTP PLTM PLTM PLSM PLLC PLMN PLDK PLMN PLDK PLQU Bolivia Bolivia a Argentina Argentin PLCM PLCM PLMK PLRR PLMK PLRR PLLA PLBR PLLA PLSS PLBR PLWB PLWB PL07 PL07 PLSQ PLLL Figure 7. Average amplitude variation ob- PLCO PLSE PLCO PLAR PLAR served from events from the (A) Japan 22.5°S 22.5°S 22.5°S 22.5°S subduction zone (JSZ), (B) European sub- PLMD PLMD duction zone (ESZ), (C) Kermadec-Tonga subduction zone (KTSZ), and (D) South Kilometers PLAN Kilometers PLAN Sandwich subduction zone (SSSZ). Peak- PLSP to-peak waveform amplitude is propor- 0 10 20 0 10 20 tional to blue circle diameter, with larger circles representing larger amplitudes at 67.5°W 67°W 66.5 °W 67.5°W 67°W 66.5°W that seismic station. Note the strong difference in amplitude distribution between 67.5°W 67°W 66.5°W 67.5°W 67°W 66.5°W events coming from the northeast/southwest direction (B and C) versus those CDfrom the northwest/southeast (A and D). Azimuths shown as black arrows pointing N N away from the source in the direction of PLCL PLCL wave propagation. The volcano is marked PLHS as a red triangle. Red ellipses highlight 22°S PL03 22°S 22°S PL03 22°S regions of distinctly smaller amplitudes. PLJR PLTP PLJR PLTP Black rectangles around arrows show PLTM PLTM the different azimuthal sets discussed in PLSM PLSM the text. Circle size scales linearly with PLLC PLDK PLLC PLDK PLMN PLMN livia normalized amplitude (maximum ampli- PLQU Bolivia PLQU Bo tina tina tude = 1), and the scale is consistent for Argen Argen PLCM PLMK PLRR PLMK PLRR all panels. PLRV PLLA PLBR PLLA PLBR PLSS PLSS PLWB PLWB PL07 PL07 PLSQ PLLL PLSQ PLLL PLSE PLSE PLCO PLCO BSSSZ PLAR PLAR 22.5°S 22.5°S 22.5°S 22.5°S PLMD

Kilometers PLAN Kilometers PLAN PLSP PLSP ASSSZ 0 10 20 0 10 20

67.5°W 67°W 66.5°W 67.5°W 67°W 66.5°W

Circle size = 1 = 0.8 = 0.6 = 0.4 = 0.2

set ASSSZ and 33° clockwise off for set BSSSZ (Fig. 7). The signals analyzed had fre- studies (e.g., Sato et al., 1989) and theoretical work (e.g., Chu et al., 2010). quencies of 0.4–1.33 Hz, with corresponding wavelengths on the order of 4–12 Velocity can be modeled independently as a first-order check; this assumes km; so no topographic effects were considered. The incoming azimuthal angle either a solid (compressibility and rigidity) or melt (compressibility only) and differences are large (Fig. 8) and suggest significant anisotropy or preferred varying percentages of each. Results are shown in Table 4 where partial melt fabric either from the subduction zone, the mantle wedge, or the crust. The dif- values are given for different thicknesses to account for the observed velocity ferences between the maximum and minimum peak-to-peak amplitudes varied reductions. Using the results of Chu et al. (2010) for a fluid-saturated porous

between an amplitude loss of 46% (KTSZ2) and 90% (SSSZ4). material (granite, rhyolite melt, water, and CO2) yields 15.2% and 14.7% veloc- Qp values (Table 5 and Table S2 [see footnote 1]) were calculated for a ity reductions for stations PLLL and PLAR, respectively, corresponding to ~8% 20 km and 25 km thick attenuating layer. The following Qp results are from the and 7% (19 and 13, if excluding the influence of water) partial melt (Table 4). 20-km-thick assumption. They follow the same spatial pattern as the ampli Laboratory studies of lherzolite and peridotite with no fluids (Sato et al., 1989) tudes, with an overall mean (excluding infinite values) of 12.4. The three show low values of Qp of 10 or less depending on pressure at high tempera- stations with consistently small Qp values (between 2.3 and 6.5, 2.9–6.7, and tures >1155°C. Combining velocity and Qp requires temperatures of 1250°C at 1.8–8.9, respectively) are PLSS, PLAR, and PLBR, all to the south and/or south- 0.5 GPa to give a partial melt of ~15%. east of the volcano. Stations PLLL and PL07, also to the S/SE of the volcano, The back-azimuth calculations (Fig. 8) showed remarkable variations be- respectively, give mean Qp values of 7.1 and 5.5, both well below the overall tween the expected azimuths (those calculated from the hypocentral locations mean. The waveforms for stations with the smallest values of Qp have under- of the events relative to the location of the volcano) and those that we deter- gone the most attenuation and therefore have the smallest signal relative to mined from the waveform data. The maximum variation was 60° (Fig. 8), and the maximum signal amplitude. the minimum was 1°. The means for the JSZ, KTSZ, ESZ, and SSSZ are –19.6°, Analyzing two phases for one event—PKIKP and PKP phases for event +19.9°, –1.2°, and +17.6° (Table 6), respectively (+ is clockwise and – is counter- JSZ2—did not show any significant differences in the distribution of ampli- clockwise). Although part of this variation can be the result of 1°–5° uncertain- tudes. The pattern was still the same, with a region of low amplitudes cen- ties in alignment between the N-S component and true north, the observed tered to the southeast of the volcano. For both phases, this region—includ- variations are on average an order of magnitude larger. The events from the ing stations PLAR, PLSS, and PLLL—contained the smallest amplitudes in NE (ESZ) and SW (KTSZ) both rotate into closer alignment with the NE trend the network. of crustal fabric (perpendicular to anisotropy), suggesting that whatever is We infer that the number of arrival time picks is insufficient to gain much causing the S-wave anisotropy is also affecting the wave transmission for the insight from a full inversion; however, relative arrivals help determine the long-period teleseismic P phases. Anisotropy studies use S waves, but they geometry and properties of the potential magma body. The accuracy of rela- reveal rock fabric that affects P waves as well. The events coming from the NW tive arrival time differences is high at a few hundredths of a second for these (JSZ) and SE (SSSZ) exhibit preferential rotations of the calculated azimuths to long-period waves. Relative residuals vary from 0 to 0.8 s. We plot normalized the NE with respect to the expected azimuths. We interpret this to be caused amplitude versus time residual for event KTSZ2 (incoming from the SW) in mainly by properties of the shallow crust above the magma body, specifically a Figure 9A. There is no trend; however, this showed four quadrants, divided by system of fluid-filled cracks oriented NNE-SSW (e.g., Leidig and Zandt, 2003). arbitrary reference lines: fast and not attenuated (upper left), slow and attenuated (lower right), fast and attenuated (upper right), and slow but not attenuated (lower left). Stations from each quadrant show spatial clustering in map view. DISCUSSION Figure 9B shows that the cluster of stations S/SE of the vent, PLBR, PLSS, PLLL, PLO7, and PLAR, all show both delayed arrival times and amplitude attenuation. We infer the presence of a magma body based on prior results by others This particular combination is consistent with the presence of partial melt (e.g., including seismic Vp/Vs ratios, Bouguer gravity anomalies, magnetotelluric and Schmeling, 1985; Sato et al., 1989; Lees, 2007), and this location S/SE of the deformation data, receiver function analysis, and the already established APMB. vent shows evidence for slow and attenuated waves for teleseisms from the Our data can best be used to place additional constraints on the geometry and JSZ, KTSZ, and SSSZ. Stations to the NW show early arrivals and moderate physical properties of the inferred magma body, including percent partial melt. attenuation (likely high fracture density; see below), whereas those SW show early arrivals and no attenuation, suggesting intact country rock. Finally, the group of stations to the NE shows time delays but no attenuation. Several ef- Teleseisms versus Regional Events fects are likely occurring (see discussion below). Similar pairs of plots for other azimuths are shown in the Supplemental Materials (see footnote 1). There are several reasons that we chose to study teleseismic events. The The combination of high attenuation (low Qp) and velocity reduction first is that the events are far away; hence the arrivals can be represented strongly suggests that partial melt is present and agrees with laboratory as plane waves. This greatly simplifies the geometry. Regional slab events,

67.5°W 67°W 66.5°W 67.5°W 67°W 66.5°W

A PLCL B

22°S PL03 22°S PLCL PLJR PLTP

PLSM 22°S 22°S PLTP PLMN Bolivia Argentina PLRR PLMK PLMN PLDK Bolivia PLLA PLBR Argentina PLCM PLWB PLMK PLRR PL07 PLSQ PLSE PLBR PLCO PLWB 22.5°S PLAR 22.5°S PL07

PLAR 22.5°S 22.5°S PLAN PLMD PLSP

Kilometers Kilometers PLAN

0 10 20 N 0 10 20 N

67.5°W 67°W 66.5°W 67.5°W 67°W 66.5°W

C D

PLCL PLCL PLHS 22°S 22°S 22°S 22°S PL03 PL03 PLTP PLTP PLJR PLJR PLSM PLSM PLLC PLLC a PLMN PLDK ia PLMN Bolivi Boliv PLQU Argentina Argentina PLCM PLMK PLRR PLRR PLLA PLSS PLBR PLSS

PL07 PLLL PLSQ PLLL PLSE PLCO PLSE PLAR 22.5°S PLAR 22.5°S 22.5°S 22.5°S

Kilometers PLAN Kilometers PLAN N PLSP 0 10 20 0 10 20 N

67.5°W 67°W 66.5°W 67.5°W 67°W 66.5°W

Figure 8. Observed versus theoretical back azimuths of selected teleseismic earthquakes from the (A) JSZ, (B) ESZ, (C) KTSZ, and (D) SSSZ. Theoretical azimuth is shown in black, and observed is in green. The volcano is marked as a red triangle.

TABLE 5. Q VALUE ESTIMATES FOR A 20-KM-THICK ATTENUATING LAYER Average Station JSZ1 JSZ2 JSZ3 JSZ4 KTSZ1KSTZ2 KTSZ3 KTSZ4 SSSZ1 SSSZ2 SSSZ3 SSSZ4 SSSZ5 station PL03 18.3 19.4 11.0 21.7 10.3 8.2 6.15.4 17.0 14.5 8.913.68.1 12.5 PL07 N.D. N.D. 6.7 13.6 4.2 5.1 3.9N.D.N.D.N.D.2.2 2.6N.D. 5.5 PLAN N.D. N.D. 7.9 5.3 17.5 14.6 7.28.8 N.D. N.D. 21.2 56.3 21.6 17.8 PLAR 4.3 6.6 3.4 6.7 4.9 5.2 4.33.2 4.44.5 4.05.5 2.9 4.6 PLBR N.D. N.D. 5.1 5.1 3.2 6.7 6.78.9 N.D. N.D. 3.64.4 1.8 5.1 PLCL N.D. N.D. 6.8 13.4 8.0 9.2 6.412.2N.D.N.D.9.8 15.7 4.3 9.5 PLCM 15.5 N.D. 9.7 N.D. 11.6 N.D.N.D.6.3 N.D. N.D. 7.8N.D.N.D. 10.2 PLCO N.D. N.D. 16.5 22.4 17.0 24.2 9.928.4N.D.N.D.6.2 8.52.5 15.1 PLDK N.D. N.D. N.D. 17.5 8.3 10.3 5.5N.D.N.D.N.D.9.2 9.47.6 9.7 PLHS N.D. N.D. N.D. N.D. N.D. N.D.N.D.11.1 N.D. N.D. N.D. N.D. N.D. 11.1 PLJR N.D. N.D. 5.6 8.7 8.1 7.1 4.42.6 N.D. N.D. 8.811.6 11.0 7.5 PLLA 23.2 24.1 50.5 17.5 17.3 11.6 6.99.3 35.3 16.4 10.3 13.8 9.1 18.9 PLLC N.D. N.D. 23.2 15.2 Inf Inf InfInf N.D. N.D. 3.88.7 5.5 11.3 PLLL 13.2 8.9 6.6 N.D. 5.0 5.8 N.D. 6.56.0 9.96.4 5.73.8 7.1 PLMD N.D. N.D. 14.2 N.D. 2.5 N.D.N.D.N.D.N.D.N.D.N.D.N.D.N.D. 8.4 PLMK 10.9 20.5 7.9 28.0 15.0 10.3 7.5N.D.90.5Inf 24.6 Inf17.7 23.3 PLMN N.D. N.D. 6.5 13.9 10.0 N.D.6.8 5.0N.D.N.D.15.619.711.9 11.2 PLQU 9.9 N.D. N.D. N.D. 8.0 5.8 N.D. 2.66.3 N.D. 7.0N.D.N.D. 5.9 PLRR Inf Inf 19.5 Inf 11.2 12.1 7.28.2 5.321.911.710.310.2 11.8 PLRV N.D. N.D. N.D. N.D. 5.5 N.D.N.D.N.D.N.D.N.D.N.D.N.D.N.D. 5.5 PLSE 48.3 25.9 Inf 32.1 14.1 9.9 7.08.5 Inf45.122.028.99.2 22.8 PLSM 9.2 11.7 4.8 9.9 12.4 7.5 4.65.1 N.D. 8.76.8 8.06.1 7.9 PLSP N.D. N.D. N.D. 13.0 14.0 9.1 5.3N.D.N.D.N.D.12.131.310.0 13.5 PLSQ 5.8 8.4 6.7 4.8 9.4 N.D.N.D.3.6 23.2 9.3Inf N.D. N.D. 8.9 PLSS 2.3 3.4 2.7 4.7 4.9 3.0 3.54.5 5.56.5 6.56.1 3.5 4.4 PLTM 29.2 53.4 11.2 N.D. 15.2 N.D.N.D.N.D.114.3 81.3 N.D. N.D. N.D. 50.8 PLTP N.D. N.D. 9.5 19.8 8.0 5.8 6.04.1 N.D. N.D. 55.4 23.3 Inf 16.5 PLWB N.D. N.D. 11.8 14.9 12.9 8.5 6.0N.D.N.D.N.D.9.7 N.D. 3.6 9.6 Average event 15.8 18.2 11.3 14.4 9.9 9.0 6.17.6 30.8 21.8 11.9 14.9 7.912.4 Note: Average values are taken over numeric values only and do not include infinite values. Abbreviations: N.D.—no data. Infinite (Inf) values denote the stations with maximum amplitudes, which we assumed to be the least affected by the attenuating source.

by contrast, produce curved wavefronts that are more difficult to model. Attenuation and Qp Values Second, the teleseismic locations are known and are independent of the uncertain local velocity model, which affects slab event locations. Third, the There are several mechanisms involved in attenuation, among them teleseisms sample a very small part of the radiation pattern, whereas radi- damped resonance or dissipation due to relative movements along grain ation pattern effects would be strong for slab events because of the closer boundaries, viscous relaxation and fluid flow in pores, and scattering at in distances and greater range of take-off angles. Fourth, the long wavelengths homogeneities (Jackson and Anderson, 1970; Richards and Menke, 1983; Sato (4–12 km) of teleseisms means they are not sensitive to small-scale (<1 km) et al., 1989; Watanabe and Sassa, 1996). Both attenuation and velocity of seis- structures. Seismograms from slab events with shorter wavelengths show mic waves change near the solidus as well as when partial melt is present, with ringing from local structures. Finally, the teleseisms are pure P waves and attenuation increasing (Qp decreasing) and velocity decreasing. This condition have nearly vertical incidence, again providing simplicity. The study of slab corresponds to the lower right part of Figure 9A (red symbols). Conversely, events would be complementary and should be done. They would poten- intact crust has low attenuation (high Qp) and higher velocities as in the upper tially include both P and S waves and may have favorable geometry and/or left of Figure 9A (blue symbols). Scattering by inhomogeneities may explain higher resolution assuming the modeling factors above are properly taken high attenuation but with higher velocities retained (e.g., Richards and Menke, into account. 1980), and further, faults and fractured zones have been identified using atten-

GEOSPHERE | Volume 13 | Number 3 Farrell et al. | Teleseisms beneath Uturuncu volcano, Bolivia Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/13/3/699/2316366/699.pdf 713 by guest on 30 September 2021 on 30 September 2021 by guest Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/13/3/699/2316366/699.pdf Research Paper Arrow denotes the azimuth of incoming wavefront. (red), late and attenuated. (B) The same data in map view; note the spatial coherence of the events from several quadrants of part A. (blue), early and not attenuated; upper right (black), late and not attenuated; lower left (green), early and attenuated, and lower right Figure 9. (A) Amplitude versus time for event KTSZ2. The grid lines are for reference and divide the field into four quadrants: Upper left 22.0 °S 22.5 °S Normalized amplitude 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9 1 B 0 A PLJR 0. 1 PLLC PLQU 67.5 °W PLS M PLA N 0. 2 PLCO PLAN PLLA PLSP PLSE PLJR PLQU PLDK 0.3 PLWB PLSE PLSM PL07 PLLA PL03 PLBR PLCL PLCO Time delay (s ) PLBR KT PLSS PLSS SZ2 PLLL PLLC 0.4 PLSP

PLMK

°W 67.0 PLMK PLCL PL07 PLTP 0.5 PLDK PLAR PLRR PL03 PLLL PLRR 0.6 PLWB PLTP 25 km 0.7

PLAR 66.5

0.8

° W

GEOSPHERE | Volume 13 | Number 3 Farrell et al. | Teleseisms beneath Uturuncu volcano, Bolivia 714 Research Paper

TABLE 6. AZIMUTHAL VARIATIONS OF TELESEISMIC ARRIVALS Event group codeFull name Azimuth from volcano to epicenter Mean azimuthal variationMaximum/minimum JSZJapan subduction zones NW –19.6+15/–62 KTSZ Kermadec-Tonga subduction zone SW +19.9+55/–19 ESZ European subduction zone NE –1.2 +28/–30 SSSZ South Sandwich subduction zone SE +17.6+50/–17 +—clockwise; –—counterclockwise.

uation tomography (Watanabe and Sassa, 1996). This corresponds to the lower Validation of the Method left of Figure 9A (green symbols). Finally, the upper right portion of Figure 9A (black symbols) shows low velocities but low attenuation. This may be caused We are aware of only a few studies of teleseismic amplitudes similar to by lithology or other causes, but the low attenuation suggests that little partial this one (e.g., Ward and Young, 1980). Therefore, we tested the efficacy of melt is present. our teleseismic amplitude method by applying it to a completely indepen- Schurr et al. (2003) performed a regional-scale study of the subduction dent region. Using the Temporary Array (TA) stations in Florida (Fig. S1 [see zone in the Central Andes. They found low values of Qp (~80) in the vicinity of footnote 1]), we can show that this method of comparing raw amplitudes of the active volcanoes to the west, but resolution near Uturuncu was poor. Low teleseismic body wave arrivals is sensitive to large-scale variations as well Qp (also ~80) was also observed beneath Cerro Tuzgle to the SSE. Similarly, as smaller-scale ones, such as those observed at our network at Uturuncu Haberland and Rietbrock (2001) determined Qp values of ~100 near the border volcano. Because the Florida stations are more geographically distributed between Bolivia and Chile (to the west and southwest of our study), which they than those in Bolivia, we first checked to determine that the Florida rays interpreted to be the magma source feeding the APVC. Our calculated values didn’t cross any nodal planes in the focal sphere or result from focusing of

of Qp are quite low, an average of 12.4 with a few single-digit values. These different waves on a travel-time curve. All the waves are Pdiff from 106° to are consistent with loss of amplitude of 46%–90% over just a few wavelengths. 112°. Additionally, we determined that the distances between source and Similar low values have been observed in the laboratory (Sato et al., 1989) stations wouldn’t result in interference from multiple contemporaneous associated with partial melting of peridotite. arrivals, which could increase waveform amplitude. In the case of Florida, The distribution of amplitudes at Uturuncu can be viewed in several differ- the peninsula is composed of relatively uniform thick Quaternary and Ter- ent ways to best identify anomalous regions. These amplitudes were shown tiary limestone units with small amounts of sandstone, whereas the north- above in Figure 7 with symbols corresponding to the direct observations. To ern region—the region showing larger amplitudes—is mainly composed of highlight the amplitude differences regardless of source azimuth, a summary Miocene siliciclastics and carbonate sediments (South Florida Information plot showing averages of the best three events from each direction (except the Access; sofia.usgs.gov). Seismic energy is retained traveling through older, ESZ) is shown in Figure 10. The same highly attenuating zone SE of Uturuncu well-consolidated rocks to the north and northwest, along the panhandle, is observed. and lost traveling through the layered rocks of the peninsula. The greater Although our calculated Qp values are quite low, values such as these are attenuation in the peninsula results from both scattering as the seismic en- not unprecedented. Schlue et al. (1996) have determined Q values for both P ergy travels through the layered rock and encountering fluids, such as the and S of ~30 at frequencies of ~1 Hz for the Socorro magma body. Similarly, numerous aquifers in the peninsula of the state. This result is strong enough Wilcock et al. (1995) determined Q values of, at minimum, 10–20 at the East to overcome the effect of distance on seismic amplitude, as the stations lo- Pacific Rise, which they attributed to thickening of a layer that is interpreted cated on the peninsula are the closest to the epicenter but show the lowest to be high-porosity lava flows and pillows (e.g., Hooft et al., 1996). Similarly, a amplitudes. Additionally, the majority of the sinkholes in Florida occur from shallow intrusion of gas-rich magma into the flank of Mount Etna exhibited Qp the middle of the state (i.e., Orlando) to the north, with few occurring on the values of 10–30 (Martínez-Arévalo et al., 2005). Wilcock et al. (1995) ascribe Q panhandle (Florida Department of Environmental Protection; dep.state.fl.us). values of 20–30 to “no more than a few percent melt” in this basaltic environ- Therefore, karst features, such as sinkholes and caves, may cause scattering ment. There are also non-magmatic examples of fluid in pore spaces bringing as well. As such, while our Uturuncu network shows large variations in an apparent Qp to single-digit values (e.g., Korneev et al., 2004). Even without par- area of diameter <100 km, the majority of the peninsula of Florida shows a tial melt, raising the temperature of basalts and gabbros to near the solidus can lack of variation in amplitudes and low overall amplitudes over a much larger give Q values of 20–40 at 10 Hz, and Q decreases from there with the introduc- area (Fig. S1 [see footnote 1]). However, these observations reveal that the tion of partial melt (e.g., Wilcock et al., 1992, 1995). Additionally, increasing the method allows determination of large-scale features that are correlated with thickness of the attenuating layer would increase our calculated values of Qp. crustal geology.

67.5°W 67.0°W

22.0°S Figure 10. Averaged values of normalized amplitudes in which large, blue circles represent lower attenuation and small, red circles greater attenuation. Averages are for nine events—the best three from each of the Japan subduction zone (JSZ1, JSZ3, and JSZ4), Kermadec-Tonga subduction zone (KTSZ1, KTSZ2, and KTSZ3), and South Sandwich subduction zone (SSSZ1, SSSZ2, and SSSZ3). Number beneath the station name specifies the number of amplitudes averaged for that station. Note erage normalized amplitude the persistent region of decreased ampli- Av tudes to the southeast of the volcano.

Circle size 22.5°S = 0.75

= 0.60

= 0.45

= 0.30

Frequency Dependence Estimates of Partial Melt

Crustal Qp for teleseismic earthquakes may show frequency dependence Many previous investigations of attenuation and partial melt have mainly in the 0.5–3.0 Hz band (Bache et al., 1986); all processing in this paper was focused on the upper mantle and mid-ocean ridge systems, as well as labora done using a filter of 0.375–1.5 Hz, therefore intersecting this band. This sug- tory studies. These benefit from the fact that upper mantle and mid-ocean gests that within this band, Q increases with increasing frequency (Morozov, ridge basalt (MORB) compositions are fairly uniform worldwide. In contrast, 2008). We see this frequency dependence when we compare the calculated volcanoes have significant structural and compositional variations, as well percentage of energy difference between the minimum and maximum ampli- as significant concentrations of fluids. Nevertheless, it is common practice to tudes for each teleseism to the period of the waveform, though it manifests extrapolate from laboratory studies to real-world situations (e.g., Sato et al., as just a slight difference (within ±7 values of Qp and decreasing to <2 with 1989). Velocity studies, particularly of Vp/Vs ratios, often rely on partial melt to

decreasing Qp). To limit this effect, estimates of Qp (Tables 5 and S2 [see foot- explain anomalies, as well as presence of fluids such as water or CO2. We did note 1]) from our attenuation data are only presented processed under a filter not take into account varying melt geometries (with the exception of assuming of 0.375–1.5 Hz, and only the dominant frequency of each waveform within interconnected melt) or dynamic melting (Li and Weidner, 2013). this band is used in calculations of Qp. This is to ensure that variations in Qp We believe that the partial melt and thickness estimates of the magma within each event are mostly attributed to the physical parameters of the crust body derived from incorporating 8 wt% water into the melt provide the best beneath Uturuncu. model for this case. While the water content may vary spatially, there is ample

evidence to suggest that crustal magmas in the APVC contain water (e.g., SW (ESZ and KTSZ, respectively; Fig. 7). This effect suggests that the waves Sparks et al., 2008; Muir et al., 2014), with Laumonier et al. (2017) claiming a arriving from different azimuths encountered different subsurface properties minimum of 8 wt%. Additionally, there is evidence of a hydrothermal system in the crust beneath Uturuncu. This could be the result of a preferential orien- at Uturuncu (Sparks et al., 2008; Jay et al., 2012), and the observed magneto- tation of fluid-filled cracks, as is commonly the case in the crust (e.g., Savage, telluric anomalies require the presence of another conducting fluid (i.e., saline 1999; Leidig and Zandt, 2003). This orientation agrees with the NW-SE align- aqueous fluid), as an unrealistic melt fraction would be necessary to result in a ment of earthquake locations (Jay et al., 2012) as well as the orientation of 3–7 Wm anomaly (Comeau et al., 2016). the Lipez-Coranzuli lineament (Lema and Choque, 1996) and the 330° upper- Comparing the estimates for partial melt and magma body thicknesses crustal anisotropy (Leidig and Zandt, 2003; Zandt et al., 2003). Thus, our tele (Table S3 [see footnote 1]) with the depths to the top of the APMB given by re- seismic data are in agreement with these independently observed trends. ceiver functions (McFarlin et al., 2014; Table S4 [see footnote 1]), we see some correlation. While the station overlying the shallowest part of the APMB, PLAN, commonly plots in the upper left quadrant (i.e., Fig. 9), stations PL07, PLAR, Teleseismic Waveform Bending PLBR, PLLL, and PLSS all overlie a shallower-than-average part of the modeled APMB. In addition, station PLAR is over one standard deviation shallower than We observe that the predicted and calculated back azimuths for the tele- the average depth to the surface of the magma body, and station PL07 is just seismic events differ (Fig. 8). This suggests that these waves interact with one slightly within one standard deviation (by 0.011) from the mean. The fact that or more strong refractors along the raypath. Some variation between theoreti these two data sets don’t completely agree on the thickness of the melt layer cal and calculated values of back azimuth may be the result of misorientation could mean that in addition to a slightly thicker APMB beneath the stations of seismometers (Koch and Kradolfer, 1999), uncertainties in earthquake loca- to the SE of the volcano, there could be a larger percent melt and/or greater tions, errors in observation at the station, errors in data processing, or geology water content. along the raypath—i.e., lateral velocity variations, as from dipping interfaces (Krüger and Weber, 1992; Lin and Roecker, 1996; Koch and Kradolfer, 1997). Seismic waves are refracted out of high-velocity slabs (Vidale, 1987), causing Directional Effects a systematic change in raypath for seismic energy that traveled through the slab. We believe that this is what gives the broad counterclockwise-clockwise The upper surface of the APMB has been modeled to be irregular (McFarlin trends in our azimuth study. Because the seismic energy is traveling from great et al., 2014); hence, we cannot rule out the possible effects of focusing and/or distance, we don’t expect the waves to be influenced by heterogeneities in defocusing of rays. Because the center of the deformation signature, located the slab, which would distort the waveforms and cause amplitude variations ~3 km to the SW of Uturuncu’s summit, may show convex-upward curvature (Sleep, 1973; Vidale, 1987; Sekiguchi, 1992). Therefore, the large variability be- as the center of a diapiric bulge of the APMB, defocusing of seismic energy tween the two azimuths expressed at each station is likely the result of interac- as the wavefront passes from the APMB to the surrounding rock is a possibil- tion with a refractor that is shallower than the subducting slab. Two candidates ity for this location (e.g., Sheriff, 1975). This would reduce the amplitude for are the base of crust and the bottom of the APMB; each would give similar stations above the bulge because of diverging rays. Station PLSS, SSE of the effects to the local rock fabric mentioned above. vent, shows strong amplitude reduction, and PLBR, SW of the vent, shows modest reduction. The data from these two stations may be affected by bending. However, none of the teleseisms studied shows similar effects at all four Gravity and Vp/Vs Tomography Data close stations PLBR, PLCM, PLMK, and PLSS; therefore, we downweight these possible effects. We compare our amplitude and arrival time observations with other geo- Our attenuation data show evidence of anisotropy in the magma body and physical data to place further constraints on possible magma bodies. Gravity vicinity, a feature also revealed in the receiver functions of Zandt et al. (2003). data, analyzed by del Potro et al. (2013), show results that correlate with the There are differences in attenuation patterns between waveforms coming from teleseismic amplitude data. The area with the largest negative gravity anom- the NW/SE and those coming from the NE/SW (Fig. 7), both in shape and mag- aly (~375 mGal, the largest area of negative density contrast) is located to the nitude. The teleseisms sourced to the NW and SE (JSZ and SSSZ, respectively) southeast of the volcano (Fig. 11) and also corresponds to areas of low seis- show smaller regions of greatly diminished amplitudes—a region to the south- mic P-wave velocity from the tomographic inversion of Kukarina et al. (2014). west of the volcano and a corridor centered on the volcano and extending in Likewise, the general shape of the negative gravity anomaly is reflected in the NW-SE direction, respectively. This differs from the broad zone of slightly the teleseismic event amplitudes for wave arrivals coming from the northeast reduced amplitudes sweeping counterclockwise around the volcano from the and southwest. Station PLAR, located within this most significant negative west through the southeast generated from teleseisms sourced to the NE and Bouguer anomaly, consistently shows the second smallest Qp values of any

68°W 67.5°W 67°W 66.5°W

AnGrav GravUt 21.5°S 21.5°S Bouguer anomaly (mGal)

PLHS PLCL 22°S PLKN 22°S PL03 PLJR PLTP PLTM PLSM PLLC PLMN PLDK PLQU PLCM # PLMK PLRR PLRV PLBR PLLA PLSS PLWB PL07 PLSQ PLLL PLSE PLCO 22.5°S PLAR 22.5°S PLMD PLTT PLAN PLSP

km 01530 N 23°S 23°S 68°W 67.5°W 67°W 66.5°W

Figure 11. Locations of Bouguer gravity anomalies in relation to the Probing Lazufre and Uturuncu TOgether: NS F, NERC, NSERC, Sergeotecmin, Sernageomin, Observatorio San Calixto, Uni- versidad Nacional de Salta, Universidad Mayor San Andres, Universidad de PotoSi, SERNAP, Chilean Seismological Service, Universidad de San Juan (PLUTONS) seismic network. Labeled large dots are PLUTONS seismic stations. All other dots show locations of gravity measurements; for explanation of AnGrav and GravUt, see del Potro et al. (2013). Legend shows the source of gravity measurements as well as the strength of the Bouguer anomaly. Solid lines and/or circles show locations of volcanoes and/or calderas (Salisbury et al., 2010). Dashed pink circles show the surface projections of the del Potro et al. (2013) diapiric ascent, and thick black dashed line shows the outline of the anomalous slow and attenuated area we interpret as partial melt. Note the large negative anomaly beneath stations PLAR, PLMD, and PLSP and the slightly smaller one near PLMK, PLCM, PLBR, and PLSS. Modified from R. del Potro (2014, written commun.).

station in the network and is also consistently delayed in time, suggesting that (Stankova et al., 2008). Unlike several of the most recent models of the surface it is both experiencing high attenuation and significant reduction in velocity. of the APMB, the SMB has a relatively flat upper surface (Rinehart et al., 1979; This same Bouguer anomaly extends, though at a reduced amplitude, beneath Balch et al., 1997). The SMB has been modeled to a depth of 19 km (Sanford stations PL07, PLBR, PLLL, and PLSS. These same stations have low seismic Qp et al., 1977), similar to the depth of the APMB at 4–25 km depth below sea level values as well as delayed times. This suggests that the attenuation, time, and (Zandt et al., 2003; Ward et al., 2014; Fig. 1). Additionally, the SMB values for gravity anomalies are similarly sourced. This relationship has also been seen Qp of ~30 (Schlue et al., 1996) fall within our range of Qp values but are higher in other volcanic terrains, such as Krafla volcano in Iceland and at Yellowstone than our smallest Qp values of as low as 1.8. This suggests a lower percent (e.g., Le Mével, 2009; Miller and Smith, 1999). partial melt in the SMB than in some, but not all, parts of the APMB. The SMB, Varying the density contrast (Dr) among end member values considering therefore, is similar to the APMB—it has been modeled to a comparable depth, realistic melt fraction values changes the size and shape of the observed grav- has comparable physical properties, and has been experiencing uplift. These ity anomaly (del Potro et al., 2013). The anomaly has a thick, wide appearance similarities may be the necessary conditions to permit the existence of long- at Dr = –70 kg/m3, which is along the liquidus for a granitoid composition. lived mid-crustal magma bodies. However, for Dr = –400 kg/m3, which is associated with 95 vol% melt, the anomaly is tighter and less distributed (del Potro et al., 2013). Therefore, the effect on seismic amplitudes would be less pronounced in the first case but Limitations more widespread, whereas the second case would result in a spatially tight, high-amplitude anomaly. Assuming that melt fraction contributes to the anom- An important large-scale limitation of this study has to do with whether or alies, it is likely that a combination of both low and moderate melt fraction not the APMB is a continuous feature (Fig. 1). This would mean that all waves contributes to the amplitude anomaly distribution that we observe. We have from teleseismic events must travel through it, and are all slowed and attenu- calculated values of percent partial melt (Table 4) that vary between 0 and 10 ated. Thus all our results are based on relative measurements. (0 and 45, if discounting the effects of water). If we look at the partial melt per- In this paper, partial melt is determined using relative travel times through centages calculated for a melt with no water, values above 35% are not likely the magma body. This makes the assumption that the fastest raypath was un- given the lack of an S-wave shadow zone (Chu et al., 2010). To account for this, affected by encountering partial melt. Therefore, the values that we calculate the anomaly would need to have a greater thickness along the raypath for the are minima. However, using relative travel-time delays highlights the differ- stations showing 35% or more, or perhaps these paths are more influenced by ences seen within the seismic network. Also, given the fact that the gravity the presence of water in the melt. Likely, it is a combination of these two fac- model of the magma structure under Uturuncu (del Potro et al., 2013) doesn’t tors—varying thickness and percent of water in melt. Given 15% partial melt, necessitate that magma underlies the entire area beneath the PLUTONS seis- the maximum thickness needed would be 16 km (24 km for dehydrated melt), mic network, this assumption may be valid. We make a similar assumption in which is within the realm of possibility given the model of Kukarina et al. (2014; our amplitude analysis—that the waveform with the greatest amplitude has

Fig. 1) and similar to the thickness of the Ward et al. (2014) model (from 4 to been unattenuated and is therefore the baseline amplitude value (A0) in our 25 km, or a maximum of 21 km thick) and the thickness values determined Qp calculations. This could mean that we are underestimating the attenua- using receiver function data (McFarlin et al., 2014). However, the mean value of tion occurring. However, given the logarithmic nature of the calculation, even

thickness given 15% partial melt is 10.2 km for hydrated melt and 15.3 km for increasing A0 by a factor of 10 resulted in very little change in the lowest Qp dehydrated. Our data favor a moderate percent of water-rich partial melt and, values (a decrease of <3 in Qp values ≤10). The greatest underestimates would therefore, a strongly attenuating body of varying thicknesses. occur for those values of Qp that are 30 and above. There were several other limitations to our study, three of which deal with the geometry of our network and of teleseismic earthquake distribution. Comparison with the Socorro Magma Body Firstly, our study was limited by the aperture of the 62–97-km-diameter seismic network, whereas the feature of interest may extend beyond the network. The results that we see for seismic waves interacting with the APMB can However, the center of the inflation source is ~3 km from the center of the net- be compared to seismic waves interacting with another large crustal magma work, and the Bouguer gravity anomaly shows significant variation within the body—the Socorro magma body (SMB) near the Rio Grande rift in New Mex- magma body beneath our network. Thus, the network straddles the strong part ico. These two mid-crustal magma bodies have several similarities despite of the anomalous zone. Secondly, the study was limited by the distribution and their very different tectonic settings, with the APMB in a zone of orogeny and number of distant large subduction zone events. The threshold between clear, crustal thickening and the SMB in a region of extensional continental rift. The unfiltered events and those needing to be filtered to clearly see the teleseism land above the SMB has been uplifting at a rate of ~2 mm/yr since 1912 (Pearse differs slightly by distance from the earthquake to the network but generally

and Fialko, 2010), exhibiting shallow seismicity and swarms at 6–7 km depth is the equivalent of Mw = 5. We analyzed deep teleseisms because we wanted

to limit the waveforms to those traveling through the asthenosphere and crust Bache, T.C., Bratt, S.R., and Bungum, H., 1986, High-frequency P-wave attenuation along five only once in the vicinity of the target volcano, therefore ensuring that varia- teleseismic paths from central Asia: Geophysical Journal of the Royal Astronomical Society, v. 85, p. 505–522, doi:10.1111/j.1365-246X.1986.tb04529.x. tions seen in the waveform resulted from the crust beneath Uturuncu. Also, Balch, R.S., Hartse, H.E., Sanford, A.R., and Lin, K., 1997, A new map of the geographic extent the epicentral distances ensure a relatively steep angle of incidence, with a of the Socorro Mid-Crustal Magma Body: Bulletin of the Seismological Society of America, range between 4° and 24° with respect to the vertical. Our study would have v. 87, p. 174–182. Cahill, T., and Isacks, B.L., 1992, Seismicity and shape of the subducted Nazca plate: Journal of benefitted from large teleseismic events from similar depths and well distrib- Geophysical Research. Solid Earth, v. 97, p. 17,503–17,529, doi:10 .1029/92JB00493. uted around the volcano. Because this is not the case, our analysis may miss Chu, R., Helmberger, D.V., Sun, D., Jackson, J.M., and Zhu, L., 2010, Mushy magma beneath or underrepresent structures that are not aligned with the azimuths of the Yellowstone: Geophysical Research Letters, v. 37, no. 1, L01306, doi:10.1029/2009GL041656. Comeau, M.J., Unsworth, M.J., Ticona, F., and Sunagua, M., 2015, Magnetotelluric images of seismic data. magma distribution beneath Volcan Uturuncu, Bolivia: Implications for magma dynamics: Geology, v. 43, p. 243–246, doi:10.1130/G36258.1. Comeau, M.J., Unsworth, M.J., and Cordell, D., 2016, New constraints on the magma distribution CONCLUSIONS and composition beneath Volcán Uturuncu and the southern Bolivian Altiplano from magnetotelluric data: Geosphere, v. 12, p. 1391–1421, doi:10.1130/GES01277.1. del Potro, R., Díez, M., Blundy, J., Camacho, A., and Gottsmann, J., 2013, Diapiric ascent of silicic Detailed analyses of waveforms from 14 teleseismic events show that the magma beneath the Bolivian Altiplano: Geophysical Research Letters, v. 40, p. 2044–2048, peak-to-peak amplitudes, zero-to-peak amplitudes, and relative travel-time doi:10.1002/grl.50493. de Silva, S.L., 1989, Altiplano-Puna volcanic complex of the Central Andes: Geology, v. 17, residuals change across the network in a way that agrees with, and better p. 1102–1106, doi:10.1130/0091-7613(1989)017<1102:APVCOT>2.3.CO;2. constrains, models of the thickening of the APMB in the vicinity of Uturuncu de Silva, S.L., Self, S., Francis, P.W., Drake, R.E., and Carlos, R.R., 1994, Effusive silicic vol volcano. There is also azimuthally dependent variability oriented in a NW-SE canism in the Central Andes: The Chao dacite and other young lavas of the Altiplano-Puna volcanic complex: Journal of Geophysical Research, v. 99, p. 17,805–17,825, doi:10 .1029 direction, suggesting significant crustal changes. We are not the first to ob- /94JB00652. serve varying values of Q encountered during different seismic raypaths under Fialko, Y., and Pearse, J., 2012, Sombrero uplift above the Altiplano-Puna magma body: Evi- Uturuncu (Jay et al., 2012); however, our results are more quantitative and dence of a ballooning mid-crustal diapir: Science, v. 338, p. 250–252, doi:10.1126/science cover a larger area. Also, the attenuation results correlate well with a Bouguer .1226358. Frankel, A., 1982, The effects of attenuation and site response on the spectra of microearth- gravity anomaly, suggesting that the source of the gravity anomaly and that of quakes in the northeastern Caribbean: Bulletin of the Seismological Society of America, the attenuation anomalies are related. v. 72, no. 4, p. 1379–1402. Our results are complementary to studies of gravity, receiver functions, Galluzzo, D., and La Rocca, M., 2013, Wavefront distortion across Mt. Vesuvius observed by a seismic array: Annals of Geophysics, v. 56, no. 4, S0448, doi:10.4401/ag-6454. magnetotellurics, and Vp/Vs tomography and other studies and provide addi- Haberland, C., and Rietbrock, A., 2001, Attenuation tomography in the western Central Andes: A tional constraints on the magma body beneath Uturuncu. Specifically, our data detailed insight into the structure of a magmatic arc: Journal of Geophysical Research. Solid suggest a magma body with <10%–20% partial melt. This magma body is 14 Earth, v. 106, p. 11,151–11,167, doi:10.1029/2000JB900472. Hashin, Z., and Shtrikman, S., 1963, A variational approach to the elastic behavior of multiphase by 34 km in size, with long axis NW-SE. It has an average thickness of 10.2 km minerals: Journal of the Mechanics and Physics of Solids, v. 11, p. 127–140, doi:10 .1016/0022 given a percent partial melt of 15% and 8 wt% water and is centered ~20 km -5096(63)90060-7. SE of Uturuncu’s summit. Henderson, S.T., and Pritchard, M.E., 2013, Decadal volcanic deformation in the Central Andes Volcanic Zone revealed by InSAR time series: Geochemistry Geophysics Geosystems, v. 14, p. 1358–1374, doi:10.1002/ggge.20074. Henderson, S.T., Pritchard, M.E., Elliott, J., del Potro, R., and Delgado, F., 2014, Observations and ACKNOWLEDGMENTS models of ground deformation from the PLUTONS Project: Lazufre and Uturuncu, Central Andes: San Francisco, California, 2014 Fall Meeting, American Geophysical Union, abstract We acknowledge the following sources of data and software: National Earthquake Information V23F-01. Center (NEIC), Boulder Real Time Technologies (BRTT), TauP, South Florida Information Access Hickey, J., Gottsmann, J., and del Potro, R., 2013, The large-scale surface uplift in the Altiplano- (sofia.usgs .gov), Florida Department of Environmental Protection (dep.state.fl.us), and the Wave- Puna region of Bolivia: A parametric study of source characteristics and crustal rheology form suite. The paper benefitted from discussions with M. West, R. del Potro, G. Thompson, using finite element analysis: Geochemistry Geophysics Geosystems, v. 14, p. 540–555, doi: J. Braunmiller, and H. McFarlin. Funding was provided by the National Science Foundation Conti- 10.1002/ggge.20057. nental Dynamics program via a subcontract to University of South Florida from the University of Hooft, E.E.E., Schouten, H., and Detrick, R.S., 1996, Constraining crustal emplacement processes Alaska Fairbanks, project #0909254. from the variation in seismic layer 2A thickness at the East Pacific Rise: Earth and Planetary Science Letters, v. 142, p. 289–309, doi:10.1016/0012-821X(96)00101-X. Jackson, D.D., and Anderson, D.L., 1970, Physical mechanisms of seismic-wave attenuation: Re- views of Geophysics, v. 8, no. 1, p. 1–63, doi:10.1029/RG008i001p00001. REFERENCES CITED Jay, J.A., Pritchard, M.E., West, M.E., Christensen, D.H., Haney, M., Minaya, E., Sunagua, M., Allmendinger, R.W., and Gubbels, T., 1996, Pure and simple shear plateau uplift, Altiplano-Puna, McNutt, S.R., and Zabala, M., 2012, Shallow seismicity, triggered seismicity, and ambient Argentina and Bolivia: Tectonophysics, v. 259, p. 1–13, doi:10.1016/0040-1951(96)00024-8. noise tomography at the long-dormant Uturuncu volcano, Bolivia: Bulletin of Volcanology, Allmendinger, R.W., Jordan, T.E., Kay, S.M., and Isacks, B.L., 1997, The evolution of the Altiplano- v. 74, p. 817–837, doi:10.1007/s00445-011-0568-7. Puna plateau of the Central Andes: Annual Review of Earth and Planetary Sciences, v. 25, Jurkevics, A., 1988, Polarization analysis of three-component array data: Bulletin of the Seismo- p. 139–174, doi:10.1146/annurev.earth.25.1.139. logical Society of America, v. 78, no. 5, p. 1725–1743.

Koch, K., and Kradolfer, U., 1997, Investigation of azimuth residuals observed at stations of Perkins, J.P., Finnegan, N.J., Henderson, S.T., and Rittenour, T.M., 2016, Topographic constraints the GSETT-3 alpha network: Bulletin of the Seismological Society of America, v. 87, no. 6, on magma accumulation below the actively uplifting Uturuncu and Lazufre volcanic centers p. 1576–1597. in the Central Andes: Geosphere, v. 12, no. 4, p. 1078–1096, doi:10.1130/GES01278.1. Koch, K., and Kradolfer, U., 1999, Determination of mislocation vectors to evaluate bias at Pritchard, M.E., and Simons, M., 2002, A satellite geodetic survey of large-scale deformation of GSETT-3 primary stations: Journal of Seismology, v. 3, no. 2, p. 139–151, doi:10 .1023/A: volcanic centres in the Central Andes: Nature, v. 418, p. 167–171, doi:10.1038/nature00872. 1009823825899. Reyes, C.G., and West, M.E., 2011, The waveform suite: A robust platform for manipulating wave- Korneev, V.A., Goloshubin, G.M., Daley, T.M., and Silin, D.B., 2004, Seismic low-frequency effects forms in MATLAB: Seismological Research Letters, v. 82, no. 1, p. 104–110, doi:10.1785/gssrl in monitoring fluid-saturated reservoirs: Geophysics, v. 69, no. 2, p. 522–532, doi:10 .1190/1 .82.1.104. .1707072. Richards, P.G., and Menke, W., 1983, The apparent attenuation of a scattering medium: Bulletin Krüger, F., and Weber, M., 1992, The effect of low-velocity sediments on the mislocation vectors of the Seismological Society of America, v. 73, no. 4, p. 1005–1021. of the GRF array: Geophysical Journal International, v. 108, no. 1, p. 387–393, doi:10.1111/j Riller, U., Petrinovic, I., Ramelow, J., Strecker, M., and Oncken, O., 2001, Late Cenozoic tectonism, .1365-246X.1992.tb00866.x. caldera and plateau formation in the Central Andes: Earth and Planetary Science Letters, Kukarina, E., West, M.E., and Koulikov, I., 2014, Seismic tomography reveals magma chamber v. 188, p. 299–311, doi:10.1016/S0012-821X(01)00333-8. location beneath Uturnco volcano (Bolivia): Vienna, Austria, 2014 European Geosciences Rinehart, E.J., Sanford, A.R., and Ward, R.M., 1979, Geographic extent and shape of an extensive Union General Assembly, abstract 5477. magma body at mid-crustal depths in the Rio Grande Rift near Socorro, New Mexico, in Laumonier, M., Gaillard, F., Muir, D., Blundy, J., and Unsworth, M., 2017, Giant magmatic water Riecker, R.E., ed., Rio Grande rift: Tectonics and Magmatism: Washington, D.C., American reservoirs at mid-crustal depth inferred from electrical conductivity and the growth of the Geophysical Union, p. 237–252, doi:10.1029/SP014p0237. continental crust: Earth and Planetary Science Letters, v. 457, p. 173–180, doi:10.1016/j.epsl Salisbury, M.J., Jicha, B.R., de Silva, S.L., Singer, B.S., Jiménez, N.C., and Ort, M.H., 2010, .2016.10.023. 40Ar/39Ar chronostratigraphy of Altiplano-Puna volcanic complex ignimbrites reveals the de- Lees, J.M., 2007, Seismic tomography of magmatic systems: Journal of Volcanology and Geo- velopment of a major magmatic province: Geological Society of America Bulletin, v. 123, thermal Research, v. 167, no. 1, p. 37–56, doi:10.1016/j.jvolgeores.2007.06.008. no. 5–6, p. 821–840, doi:10.1130/B30280.1. Leidig, M., and Zandt, G., 2003, Modeling of highly anisotropic crust and application to the Alti- Sanford, A.R., Mott, R.P., Shuleski, P.J., Rinehart, E.J., Caravella, F.J., Ward, R.M., and Wallace, plano-Puna volcanic complex of the Central Andes: Journal of Geophysical Research, v. 108, T.C., 1977, Geophysical evidence for a magma body in the crust in the vicinity of Socorro, p. ESE 5-1–ESE 5-15, doi:10.1029/2001JB000649. New Mexico: The Earth’s Crust, v. 20, p. 385–403, doi:10.1029/GM020p0385. Lema, J.C., and Choque, N., 1996, Carta Geológica Nacional: Servicio Nacional de Geología y Sato, H., Sacks, I.S., Murase, T., Muncill, G., and Fukuyama, H., 1989, Qp-melting temperature Minería Servicio Geológico de Bolivia series I-CGB, scale 1:100,000. relation in peridotite at high pressure and temperature: Attenuation mechanism and impli- Le Mével, H., 2009, Relationship between seismic and gravity anomalies at Krafla volcano, North cations for the mechanical properties of the upper mantle: Journal of Geophysical Research. Iceland (B.S. research project): University of Iceland, 33 p., http://hdl .handle.net/1946/3201 Solid Earth, v. 94, p. 10,647–10,661, doi:10.1029/JB094iB08p10647. (accessed April 2015). Savage, M.K., 1999, Seismic anisotropy and mantle deformation: What have we learned from Li, L., and Weidner, D.J., 2013, Effect of dynamic melting on acoustic velocities in a partially shear wave splitting?: Reviews of Geophysics, v. 37, no. 1, p. 65–106, doi:10 .1029/98RG02075. molten peridotite: Physics of the Earth and Planetary Interiors, v. 222, p. 1–7, doi:10 .1016/j Schilling, F.R., and Partzsch, G.M., 2001, Quantifying partial melt fraction in the crust beneath .pepi.2013.06.009. the Central Andes and the Tibetan Plateau: Physics and Chemistry of the Earth. Part A: Solid Lin, C.H., and Roecker. S.W., 1996, P-wave backazimuth anomalies observed by a small-aperture Earth and Geodesy, v. 26, no. 4, p. 239–246, doi:10.1016/S1464-1895(01)00051-5. seismic array at Pinyon Flat, southern California: Implications for structure and source loca- Schilling, F.R., Partzsch, G.M., Brasse, H., and Schwarz, G., 1997, Partial melting below the mag- tion: Bulletin of the Seismological Society of America, v. 86, no. 2, p. 470–476. matic arc in the Central Andes deduced from geoelectromagnetic field experiments and Martínez-Arévalo, C., Patanè, D., Rietbrock, A., and Ibánez, J.M., 2005, The intrusive process laboratory data: Physics of the Earth and Planetary Interiors, v. 103, no. 1, p. 17–31, doi:10 leading to the Mt. Etna 2001 flank eruption: Constraints from 3-D attenuation tomography: .1016/S0031-9201(97)00011-3. Geophysical Research Letters, v. 32, no. 21, doi:10.1029/2005GL023736. Schilling, F.R., Trumbull, R.B., Brasse, H., Haberland, C., Asch, G., Bruhn, D., Mai, K., Haak, V., Giese, Mavko, G., Mukerji, T., and Dvorkin, J., 2009, The rock physics handbook: Tools for seismic P., Muñoz, M., and Ramelow, J., 2006, Partial melting in the Central Andean crust: A review analysis of porous media: Cambridge, UK, Cambridge University Press, 169 p, doi:10.1017 of geophysical, petrophysical, and petrologic evidence, in Oncken, O., Chong, G., Franz, G., /CBO9780511626753. Giese, P., Götze, H.-J., Ramos, V.A., Strecker, M.R., and Wigger, P., eds., The Andes: Active Sub- McFarlin, H.L., Christensen, D.H., Thompson, G., McNutt, S.R., Ryan, J.C., Ward, K.M., Zandt, G., duction Orogeny: Berlin, Heidelberg, Springer, p. 459–474, doi:10 .1007 /978 -3 -540 -48684 -8_22 . and West, M.E., 2014, Receiver function analysis of Uturuncu volcano, Bolivia, and Lastarria/ Schlue, J.W., Aster, R.C., and Meyer, R.P., 1996, A lower crustal extension to a midcrustal magma Cordon del Azufre volcanoes, Chile: San Francisco, California, 2014 Fall Meeting, American body in the Rio Grande Rift, New Mexico: Journal of Geophysical Research, v. 101, p. 25,283– Geophysical Union, abstract V31E-4792. 25,291, doi:10.1029/96JB02464. Miller, D.S., and Smith, R.B., 1999, P and S velocity structure of the Yellowstone volcanic field Schmeling, H., 1985, Numerical models on the influence of partial melt on elastic, anelastic and from local earthquake and controlled-source tomography: Journal of Geophysical Research. electric properties of rocks. Part I: Elasticity and anelasticity: Physics of the Earth and Plane- Solid Earth, v. 104, p. 15105–15121, doi:10.1029/1998JB900095. tary Interiors, v. 41, no. 1, p. 34–57, doi:10.1016/0031-9201(85)90100-1. Morozov, I.B., 2008, Geometrical attenuation, frequency dependence of Q, and the absorption Schmitz, M., Heinsohn, W.D., and Schilling, F.R., 1997, Seismic, gravity and petrological evidence band problem: Geophysical Journal International, v. 175, no. 1, p. 239–252, doi:10.1111/j.1365 for partial melt beneath the thickened central Andean crust (21–23°S): Tectonophysics, -246X.2008.03888.x. v. 270, no. 3, p. 313–326, doi:10.1016/S0040-1951(96)00217-X. Muir, D.D., Blundy, J.D., Rust, A.C., and Hickey, J., 2014, Experimental constraints on dacite Schurr, B., Asch, G., Rietbrock, A., Trumbull, R., and Haberland, C.H., 2003, Complex patterns of fluid preeruptive magma storage conditions beneath Uturuncu volcano: Journal of Petrology, and melt transport in the central Andean subduction zone revealed by attenuation tomography: v. 55, p. 749–767, doi:10.1093/petrology/egu005. Earth and Planetary Science Letters, v. 215, no. 1, p. 105–119, doi:10 .1016 /S0012 -821X (03)00441 -2. Nakajima, J., Matsuzawa, T., Hasegawa, A., and Zhao, D., 2001, Three-dimensional struc- Sekiguchi, S., 1992, Amplitude distribution of seismic waves for laterally heterogeneous structure of Vp, Vs, and Vp/Vs beneath northeastern Japan: Implications for arc magma- tures including a subducting slab: Geophysical Journal International, v. 111, no. 3, p. 448– tism and fluids: Journal of Geophysical Research, v. 106, p. 21,843–21,857, doi:10 .1029 464, doi:10.1111/j.1365-246X.1992.tb02103.x. /2000JB000008. Sheriff, R.E., 1975, Factors affecting seismic amplitudes: Geophysical Prospecting, v. 23, no. 1, Pearse, J., and Fialko, Y., 2010, Mechanics of active magmatic intraplating in the Rio Grande Rift p. 125–138, doi:10.1111/j.1365-2478.1975.tb00685.x. near Socorro, New Mexico: Journal of Geophysical Research. Solid Earth, v. 115, p. 1–16, Sleep, N.H., 1973, Teleseismic P-wave transmission through slabs: Bulletin of the Seismological doi:10.1029/2009JB006592. Society of America, v. 63, no. 4, p. 1349–1373.

Sparks, R.S.J., Folkes, C.B., Humphreys, M.C.S., Barfod, D.N., Clavero, J., Sunagua, M.C., McNutt, Watanabe, T., and Sassa, K., 1996, Seismic attenuation tomography and its application to rock S.R., and Pritchard, M.E., 2008, Uturuncu Volcano, Bolivia: Volcanic unrest due to mid-crustal mass evaluation: International Journal of Rock Mechanics and Mining Sciences & Geome- magma intrusion: American Journal of Science, v. 308, p. 727–769, doi:10 .2475 /06 .2008 .01 . chanics Abstracts, v. 33, no. 5, p. 467–477, doi:10.1016/0148-9062(96)00005-8. Stankova, J., Bilek, S.L., Rowe, C.A., and Aster, R.C., 2008, Characteristics of the October 2005 Whitman, D., Isacks, B.L., and Kay, S.M., 1996, Lithospheric structure and along-strike segmen- microearthquake swarm and reactivation of similar event seismic swarms over decadal time tation of the central Andean Plateau: Topography, tectonics, and timing: Tectonophysics, periods near Socorro, New Mexico: Bulletin of the Seismological Society of America, v. 98, v. 259, p. 29–40, doi:10.1016/0040-1951(95)00130-1. p. 93–105, doi:10.1785/0120070108. Wilcock, W.S.D., Solomon, S.C., Purdy, G.M., and Toomey, D.R., 1992, The seismic attenuation Steck, L.K., and Prothero, W.A., 1993, Observations of direct P-wave slowness and azimuth structure of a fast-spreading mid-ocean ridge: Science, v. 258, p. 1470–1474, doi:10.1126 anomalies for teleseisms recorded in Long Valley Caldera, California: Bulletin of the Seis- /science.258.5087.1470. mological Society of America, v. 83, no. 5, p. 1391–1419. Wilcock, W.S.D., Solomon, S.C., Purdy, G.M., and Toomey, D.R., 1995, Seismic attenuation struc- Vidale, J.E., 1987, Waveform effects of a high-velocity, subducted slab: Geophysical Research ture of the East Pacific Rise near 9°30′N: Journal of Geophysical Research, v. 100, p. 24,147– Letters, v. 14, no. 5, p. 542–545, doi:10.1029/GL014i005p00542. 24,165, doi:10.1029/95JB02280. Ward, K.M., Porter, R.C., Zandt, G., Beck, S.L., Wagner, L.S., Minaya, E., and Tavera, H., 2013, Yeguas, A.G., Almendros, J., Abella, R., and Ibáñez, J.M., 2011, Quantitative analysis of seis- Ambient noise tomography across the Central Andes: Geophysical Journal International, mic wave propagation anomalies in azimuth and apparent slowness at Deception Island v. 194, p. 1559–1573, doi:10.1093/gji/ggt166. volcano (Antarctica) using seismic arrays: Geophysical Journal International, v. 184, no. 2, Ward, K.M., Zandt, G., Beck, S., Christensen, D.H., and McFarlin, H., 2014, Seismic imaging of p. 801–815, doi:10.1111/j.1365-246X.2010.04864.x. the magmatic underpinnings beneath the Altiplano-Puna volcanic complex from the joint Yuan, X., Sobolev, S., Kind, R., Oncken, O., Bock, G., Asch, G., Schurr, B., Graeber, F., Rudloff, A., inversion of surface wave dispersion and receiver functions: Earth and Planetary Science and Hanka, W., 2000, Subduction and collision processes in the Central Andes constrained Letters, v. 404, p. 43–53, doi:10.1016/j.epsl.2014.07.022. by converted seismic phases: Nature, v. 408, p. 958–961, doi:10.1038/35050073. Ward, R.W., and Young, C.Y., 1980, Mapping seismic attenuation within geothermal systems Zandt, G., Leidig, M., Chmielowski, J., Baumont, D., and Yuan, X., 2003, Seismic detection and using teleseisms with application to the Geysers–Clear Lake region: Journal of Geophysical characterization of the Altiplano-Puna magma body, Central Andes: Pure and Applied Geo- Research. Solid Earth, v. 85, p. 5227–5236, doi:10.1029/JB085iB10p05227. physics, v. 160, p. 789–807, doi:10.1007/PL00012557.

Erratum to this article Seismic attenuation, time delays, and raypath bending of teleseisms beneath Uturuncu volcano, Bolivia Alexandra K. Farrell, Stephen R. McNutt, and Glenn Thompson (first published on 30 March 2017, doi:10.1130/GES01354.1)

When this article was originally published, one author was inadvertently left out of the author list. Glenn Thompson has been added to the author list on the first page of the paper.