https://doi.org/10.1130/G47127.1 Manuscript received 20 August 2019 Revised manuscript received 23 October 2019 Manuscript accepted 9 November 2019 © 2019 Geological Society of America. For permission to copy, contact [email protected]. Magma accumulation beneath Santorini volcano, Greece, from P-wave tomography B.G. McVey1, E.E.E. Hooft1, B.A. Heath1, D.R. Toomey1, M. Paulatto2, J.V. Morgan2, P. Nomikou3 and C.B. Papazachos4 1 Department of Earth Sciences, University of Oregon, Eugene, Oregon 97403, USA 2 Department of Earth Science and Engineering, Imperial College London, London SW7 2AZ, UK 3 Department of Geology and Geoenvironment, National and Kapodistrian University of Athens, Athens 157 72, Greece 4 Geophysical Laboratory, Aristotle University of Thessaloniki, Thessaloniki 541 24, Greece ABSTRACT There is multi-disciplinary evidence for Despite multidisciplinary evidence for crustal magma accumulation below Santorini vol- shallow magma accumulation below Santori- cano, Greece, the structure and melt content of the shallow magmatic system remain poorly ni, but to date, no clear structural constraint on constrained. We use three-dimensional (3-D) velocity models from tomographic inversions the magmatic system. In 2015, the PROTEUS of active-source seismic P-wave travel times to identify a pronounced low-velocity anomaly (Plumbing Reservoirs of the Earth under Santo- (–21%) from 2.8 km to 5 km depth localized below the northern caldera basin. This anomaly rini) experiment collected dense seismic data us- is consistent with depth estimates of pre-eruptive storage and a recent infation episode, sup- ing ∼150 seismometers and ∼14,000 controlled porting the interpretation of a shallow magma body that causes seismic attenuation and ray marine sources (Fig. 1) to image the magmatic bending. A suite of synthetic tests shows that the geometry is well recovered while a range system beneath Santorini. Data from the experi- of melt contents (4%–13% to fully molten) are allowable. A thin mush region (2%–7% to ment revealed a low-velocity body below the 3%–10% melt) extends from the main magma body toward the northeast, observed as low northern caldera basin centered at 1.6 km depth, velocities confned by tectono-magmatic lineaments. This anomaly terminates northwest of attributed to a high-porosity column created by Kolumbo; little to no melt underlies the seamount from 3 to 5 km depth. These structural collapse of a limited area of the caldera foor constraints suggest that crustal extension and edifce loads control the geometry of magma (Hooft et al., 2019). The longer-offset data in the accumulation and emphasize that the shallow crust remains conducive to melt storage shortly present study deepen model coverage to 6 km to after a caldera-forming eruption. investigate the volume, geometry, and minimum melt content of shallow magma storage below INTRODUCTION of at least nine Plinian eruptions separated by Santorini. By combining geologic observations After a caldera-forming eruption, the shallow inter-Plinian periods of frequent, effusive shield- and tomographic images, we obtain a better un- magmatic system is rejuvenated by fresh injec- building events (Druitt et al., 1999). The Late derstanding of both the nature of shallow res- tions of melt from the lower crust. Some studies Bronze Age (LBA or Minoan) eruption in 1630 ervoirs at active caldera-forming arc volcanoes suggest that shallow melt accumulates where BCE was the most recent caldera-forming erup- during shield-building periods and the processes crustal stresses are least compressive due to top- tion, of ∼60–70 km3 dense-rock equivalent of that infuence the restructuring of the magmatic ographic loads and regional tectonic extension magma (Johnston et al., 2014; Karátson et al., system. (Corbi et al., 2015). How quickly the magmatic 2018). Subsequent effusive activity has formed system reforms, in what way the melt accumu- the 400-m-tall Kameni shield islands within TRAVEL-TIME TOMOGRAPHY lates, and how long shallow melt is maintained Santorini caldera, totaling ∼3.2 km3 since 197 To image the structure of the caldera to ∼6 km in the upper crust are still debated (Cooper and BCE, and most recently erupting in 1950 CE depth, we added ∼30,000 longer-offset crustal- Kent, 2014; Barboni et al., 2016; Cashman et al., (Nomikou et al., 2014). From 2011 to 2012, San- refraction, Pg, arrival picks that cross the caldera 2017; Rubin et al., 2017). torini experienced an unrest period including to the existing ∼200,000 Pg picks of Heath et al. Santorini, Greece, is an ideal volcano to ad- ground deformation measured by InSAR and (2019) (see the GSA Data Repository 1). Seis- dress competing ideas of shallow magma stor- GPS that was modeled to suggest 0.01–0.02 km3 mic waves that travel below the northern caldera age because its geologic and historic eruption of infation between 4 and 5 km depth northwest basin are attenuated with delayed arrivals—as history is well studied and it currently main- of the Kameni islands (Parks et al., 2015). This expected for magma bodies (Lees, 2007). We tains a shallow magma body which has sourced infation was attributed to an intrusion of melt used azimuthal swaths to enhance lateral con- small eruptions following a recent caldera col- that pooled at the same depth as that constrained tinuity for rays passing through the magmatic lapse. Santorini is a semi-submerged volcano for pre-eruptive storage of Kameni island lavas system (Fig. 2). To improve our ability to pick in the Hellenic volcanic arc that has a history by a melt inclusion study (Druitt et al., 2016). the attenuated caldera-crossing data, we stacked 1GSA Data Repository item 2020061, descriptions and illustrations of the seismic data analysis, tomographic inversion, physical property analysis, synthetic tomographic modeling, and volume calculation, is available online at http://www.geosociety.org/datarepository/2020/, or on request from [email protected]. CITATION: McVey, B.G., et al., 2020, Magma accumulation beneath Santorini volcano, Greece, from P-wave tomography: Geology, v. 48, p. XXX–XXX, https:// doi.org/10.1130/G47127.1 Geological Society of America | GEOLOGY | Volume XX | Number XX | www.gsapubs.org 1 Downloaded from https://pubs.geoscienceworld.org/gsa/geology/article-pdf/doi/10.1130/G47127.1/4902160/g47127.pdf by University of Oregon user on 15 January 2020 Heath et al. (2019) and presented a 28 × 33 × 6 km subset of the full velocity model, which has a root- mean-squared misft of 15.5 ms. The derivative weight sum (Toomey et al., 1990) and checkerboard tests (Figs. DR5–DR6) show ray coverage to 5.5–6 km depth and recovery of features with dimensions 3 × 3 × 1 km to 4 km depth and with dimensions of 5 × 5 × 2 km to 5 km depth. RESULTS The traveltime data added in this study deepen the model coverage (Fig. 3), revealing a low-velocity volume centered below the north- ern caldera basin (caldera low-velocity volume; CLV) with a limb that extends ∼15 km north- eastward (linear low-velocity volume; LLV). Ray bending provides direct evidence for the CLV. We simplify the wavefront, which has a f- nite sensitivity kernel, as a ray path that takes the fastest path from the airgun to the seismometer. Rays traveling through the starting model have evenly distributed paths between source and receiver with little bending (Fig. 2; Fig. DR2). Figure 1. Map of area of PROTEUS (Plumbing Reservoirs of the Earth under Santorini) experi- ment, Santorini volcano, Greece, showing bathymetry (Nomikou et al., 2012; Hooft et al., 2017) As the inversion progresses through each it- and seismic experiment geometry (Heath et al., 2019). Seismic recorders (in yellow) include eration, the rays bend around a volume in the ocean bottom seismometers (OBS) and land stations. Shot lines of seismic sources (in black) northern caldera basin, where geodetic models were generated from airguns on the R/V Marcus G. Langseth. Area of this study is within the suggest the recent infation was located (Parks red square. Inset map indicates regional geography and volcanic arc. et al., 2015)—clear evidence for a CLV in the northern part of the caldera. However, because the hydrophone and vertical components of the three-dimensional (3-D) velocity model (Toomey very few rays probe the CLV in the fnal model, seismometer and used a low-frequency bandpass et al., 1994) (see the Data Repository). Bathym- the recovered velocity image provides only the flter (Fig. DR1 in the Data Repository). etry was explicitly included, and the magnitude minimum perturbation required to exclude rays. We used a seismic tomography method to and roughness of model perturbations were pe- The CLV northwest of the Kameni islands invert frst-arriving P-wave traveltimes for a nalized. We followed the parameterization of is directly below the hypothesized caldera col- lapse column (Hooft et al., 2019) and has a -0.2 maximum P-wave velocity anomaly of −21% A 0 ) at 3.4 km depth (Fig. DR3). This anomaly is of s ( 0.2 e similar depth as, but of larger magnitude than, m i 0.4 t l those associated with upper crustal magma stor- e 0.6 v a age at Mount St. Helens (Washington, USA), r 0.8 T Newberry (Oregon, USA), Deception Island 1.0 Trace azimuth sort order (South Shetland Islands, Antarctica), Montser- rat (Lesser Antilles), and Avacha (Kamchatka Initial Raypaths Final Raypaths B C Peninsula, Russia) (Kiser et al., 2018; Heath 0 km 5 km 10 km 0 km 5 km 10 km et al., 2015; Ben-Zvi et al., 2009; Paulatto et al., 36.5°N 36.5°N 2019, Bushenkova et al., 2019). Surrounding the CLV to the northwest and southeast are regions of high velocity (Fig.