Volume 124, Issue 10 p. 10610-10629
Research Article
Free Access

Tectonism and Its Relation to Magmatism Around Santorini Volcano From Upper Crustal P Wave Velocity

B. A. Heath

Corresponding Author

B. A. Heath

Department of Earth Sciences, University of Oregon, Eugene, OR, USA

Correspondence to: B. A. Heath,

[email protected]

Search for more papers by this author
E. E. E. Hooft

E. E. E. Hooft

Department of Earth Sciences, University of Oregon, Eugene, OR, USA

Search for more papers by this author
D. R. Toomey

D. R. Toomey

Department of Earth Sciences, University of Oregon, Eugene, OR, USA

Search for more papers by this author
C. B. Papazachos

C. B. Papazachos

Geophysical Laboratory, Aristotle University of Thessaloniki, Thessaloniki, Greece

Search for more papers by this author
P. Nomikou

P. Nomikou

Department of Geology and Geoenvironment, National and Kapodistrian University of Athens, Athens, Greece

Search for more papers by this author
M. Paulatto

M. Paulatto

Department of Earth Science and Engineering, Imperial College London, London, UK

Search for more papers by this author
J. V. Morgan

J. V. Morgan

Department of Earth Science and Engineering, Imperial College London, London, UK

Search for more papers by this author
M. R. Warner

M. R. Warner

Department of Earth Science and Engineering, Imperial College London, London, UK

Search for more papers by this author
First published: 09 August 2019
Citations: 20


At extensional volcanic arcs, faulting often acts to localize magmatism. Santorini is located on the extended continental crust of the Aegean microplate and is one of the most active volcanoes of the Hellenic arc, but the relationship between tectonism and magmatism remains poorly constrained. As part of the Plumbing Reservoirs Of The Earth Under Santorini experiment, seismic data were acquired across the Santorini caldera and the surrounding region using a dense amphibious array of >14,300 marine sound sources and 156 short-period seismometers, covering an area 120 km by 45 km. Here a P wave velocity model of the shallow, upper-crustal structure (<3-km depth), obtained using travel time tomography, is used to delineate fault zones, sedimentary basins, and tectono-magmatic lineaments. Our interpretation of tectonic boundaries and regional faults are consistent with prior geophysical studies, including the location of basin margins and E-W oriented basement faults within the Christiana Basin west of Santorini. Reduced seismic velocities within the basement east of Santorini, near the Anydros and Anafi Basins, are coincident with a region of extensive NE-SW faulting and active seismicity. The structural differences between the eastern and western sides of Santorini are in agreement with previously proposed models of regional tectonic evolution. Additionally, we find that regional magmatism has been localized in NE-SW trending basin-like structures that connect the Christiana, Santorini, and Kolumbo volcanic centers. At Santorini itself, we find that magmatism has been localized along NE-SW trending lineaments that are subparallel to dikes, active faults, and regional volcanic chains. These results show strong interaction between magmatism and active deformation.

Key Points

  • P wave tomography of the upper crust (<3 km) around Santorini delineates basin structures including buried margins and basement faults
  • These structures, also seen in seismic reflection images, result from both the integrated tectonic history, as well as ongoing deformation
  • Tectonic structures and stresses localize volcanism in regional NE-SW basins, and, within Santorini caldera, along local NE-SW lineaments

1 Introduction

In extensional volcanic arcs, the crust is often composed of a patchwork of interacting faults that control the localization of magmatism and development of sedimentary basins. Analogue models have shown that magma may be localized within basins in different extensional regimes (Corti et al., 2003). In turn magma bodies can alter the stress state of the crust. In addition to weakening the crust through thermal effects, magmatism can facilitate the formation of low-angle faults (Parsons & Thompson, 1993), inhibit the formation of throughgoing normal faults (e.g., Faulds & Varga, 1998), and accommodate extensional strain.

Santorini is one of the most active volcanoes in the Hellenic Volcanic arc (e.g., Le Pichon & Angelier, 1979; Figure 1), resulting from the subduction of the African plate under the Aegean microplate. Regional extension, since the Oligocene-Miocene, has been accommodated both on older E-W oriented faults and on younger NE-SW oriented normal faults that remain tectonically active to present day. As a consequence of this extension, the crust has been thinned and multiple fault systems of differing orientations intersect Santorini (Figures 1 and 2). The link between tectonics and magmatism across Santorini, however, is currently not well constrained, in part because many of these features are buried under sedimentary and volcanic deposits.

Details are in the caption following the image
Tectonic setting of the Aegean. Topographic/bathymetric map of the broader Aegean/Hellenic subduction zone (bold grey line with triangles) where the African plate subducts under the Aegean microplate showing the volcanic arc (red region) and main volcanic centers (stars and bold names). The study area is indicated with a dashed box. Faults in the overriding Aegean microplate (black lines; after Caputo et al., 2012) result from subduction rollback and extrusion along the North Anatolia Fault. The Santorini study area is located between the Cyclades to the northwest and the Cretan Basin to the south, within an actively faulted region. The GPS-derived velocity vectors relative to Eurasia (black arrows) reflect differential motion between regions east and west of Santorini (Reilinger et al., 2010).
Details are in the caption following the image
The Santorini region showing the PROTEUS active seismic experiment and the main morphotectonic features. (a) Topographic/bathymetric map of Santorini and the surrounding islands showing seafloor and island seismic stations (yellow circles are stations used in inversion, grey circles are stations that were not used) and the seismic shot profiles (small red circles). Data from seismic lines recorded on stations 148 and 105 (blue/purple) are shown in Figure 3. (b) Topographic/bathymetric map with the main basins, horsts/ridges, and faults (red lines; from Hooft et al., 2017). Abreviations used are CB = Christiana Basin, CR = Christiana Ridge NC = Northern Caldera, SC= Southern Caldera, IF = Ios Fault, AH = Anydros Horst, AB = Anydros Basin, AFB = Anafi Basin, SAB = Santorini-Anafi Basin, AAB = Anafi-Amorgos Basin. Kolumbo and Kameni lines (KL1 and KL2, respectively) are after Pfeiffer (2001).

In this study we use seismic data, acquired in the marine-land active-source Plumbing Reservoirs Of The Earth Under Santorini seismic experiment, to obtain a tomographic P wave velocity model of the upper-crustal structure across Santorini volcano and the surrounding region. Seismic velocity is sensitive to fracturing, porosity, and composition, and here we use it to delineate the structure and evolution of faults, extensional basins, and volcanic features in the Santorini region. We draw on our tomographic images, and results from previous seismic reflection and other studies, to delineate basins and faults. The combined data sets, as well as results from prior studies, are used to investigate differences between the upper crust to the west and east of Santorini, build a geologic model of fault evolution, and explore the relationship between tectonism and magmatism.

2 Background

2.1 Tectonic Background

The Aegean has undergone multiple episodes of extension from at least the Oligocene-Miocene to present that were driven by slab rollback of the Hellenic subduction zone (Le Pichon & Angelier, 1979), westward extrusion of Anatolia (McKenzie, 1972), and gravitational collapse of continental crust (McKenzie, 1972). These extensional events have led to thinned crust (Makris, 1976, 1978), exhumed and rotated metamorphic blocks (Walcott & White, 1998, and others), and localized deformation along zones of crustal weakness (Jolivet et al., 2013, and citations therein), creating a variety of deforming terrains.

Geological studies show that the Cyclades (Figure 1) were extended, thinned, and rotated throughout the Miocene, but have behaved as a single rotating block since the late Miocene-early Pliocene (Walcott & White, 1998) and currently exhibit little internal deformation. Santorini and the surrounding region lies at the southeastern margin of this Cycladic block (Le Pichon & Kreemer, 2010; Figure 1).

During the Miocene to Pliocene, E-W striking normal faults and related basins formed under ~N-S extension (Anastasakis & Piper, 2005; Piper et al., 2007). One of these basins is the Christiana Basin in the western part of the study area. The Christiana Basin has been mapped using seismic reflection imaging (Tsampouraki-Kraounaki & Sakellariou, 2018; Piper et al., 2007, and citations therein) and is thought to have been truncated against N-S striking transfer faults (Piper & Perissoratis, 2003). This E-W striking basin is filled with Messinian and younger sediments (Piper & Perissoratis, 2003; Tsampouraki-Kraounaki & Sakellariou, 2018). The main faults bounding the Christiana Basin ceased activity ~1 Ma during the Pleistocene (Piper & Perissoratis, 2003) but some activity continues to present day as evidenced by small-offset, active faults observed in seismic reflection data (Tsampouraki-Kraounaki & Sakellariou, 2018).

In the latest Pliocene or early Pleistocene, the regional stress direction rotated resulting in NE-SW trending faults and subsequent basin inversion (Piper et al., 2007; Piper & Perissoratis, 2003) as well as the formation of new basins such as the Anydros and Anafi Basins on the eastern side of our study area (e.g., Hübscher et al., 2015; Nomikou et al., 2016). Following the change in faulting orientation, volcanism in the region began (Perissoratis, 1995), with initial volcanism sourced at Christiana and Akrotiri (SW Santorini; Piper et al., 2007).

Faults striking NE-SW continue to be active to present day (e.g., Bohnhoff et al., 2006; Dimitriadis et al., 2009) and are thought to be predominately extensional to trans-tensional (Hooft et al., 2017; Hübscher et al., 2015; Nomikou et al., 2016; Nomikou et al., 2018), but a component of strike-slip behavior is observed (Sakellariou et al., 2010). The NE-SW striking Santorini-Amorgos fault zone (e.g., Stiros et al., 1994), which includes the Anydros Region on the eastern side of the study area (Figure 2), is one of the most active in the Aegean and regionally is the margin between the seismically quiet Cycladic block and the diffuse seismicity observed in the eastern Aegean (Bohnhoff et al., 2006). This is consistent with GPS studies showing that the eastern Aegean moves to the SE relative to the Cylcadic block (McClusky et al., 2000; Reilinger et al., 2010).

Metamorphosed sediments of the Attico-Cycladic complex, which were exhumed during Miocene and younger extension (Piper et al., 2007; Piper & Perissoratis, 2003; Xypolias et al., 2010), highlight the role of tectonics in the evolution of the southern Aegean. These exhumed rocks compose most of the regional islands (e.g., Ios, Amorgos; Lister et al., 1984; Forster & Lister, 1999), including a portion of SW Santorini itself (Heiken & Mccoy, 1984). Moreover, these metamorphic fault blocks often extend as horsts into the Aegean Sea, as seen in seafloor bathymetry and seismic reflection profile images (Hooft et al., 2017; Hübscher et al., 2015; Nomikou et al., 2016, 2018; Perissoratis, 1995). Metamorphic basement lithics, found in the erupted ignimbrite products from Santorini, reveal that the metamorphic complex must underlie Santorini's caldera (Druitt, 2014), and it is thought that Santorini sits on a NE-SW oriented fracture network (Budetta et al., 1984).

2.2 Volcanic Background

Regional volcanism has been localized in three areas: Christiana Island, Santorini volcano, and the Kolumbo volcanic chain, arranged roughly in a NE-SW direction (Nomikou et al., 2013; Figure 2). The oldest of the regional volcanic deposits results from Christiana volcanism (active during late Pliocene) and from early volcanic centers at the SW edge of Santorini near Akrotiri (~0.6 Ma; Piper et al., 2007). Volcanism at Santorini itself began roughly 650 ka (Druitt et al., 1999), with explosive volcanism beginning around 360 ka. At least four major caldera-forming eruptions have occurred at Santorini (Druitt, 2014), the most recent of which, the 30–86-km3 (dense rock equivalent) Late Bronze Age (LBA) “Minoan” eruption, occurred 3.4 ka (Druitt, 2014, and citations therein; Johnston et al., 2014; Friedrich et al., 2006).

Post-LBA effusive eruptive vents are located along the Kameni line, a NE-SW trending line of vents in the caldera (Nomikou et al., 2014; Pyle & Elliott, 2006) that results from either a deep-seated fault focusing magmatism, a caldera ring fault, or a region of diking (Konstantinou et al., 2013; Newman et al., 2012; Papadimitriou et al., 2015; Saltogianni et al., 2014; Tassi et al., 2013). The Kolumbo line, a similarly oriented lineament north of the Kameni line (see Figure 2), is observed as a linear alignment of older volcanic centers in the NE caldera and is thought to strike toward the Kolumbo volcanic chain NE of Santorini. The Kolumbo volcanic chain is a NE striking alignment of submarine cones (e.g., Hooft et al., 2017; Nomikou et al., 2012, 2013), the largest of which is Kolumbo Seamount (Hübscher et al., 2015; Nomikou et al., 2016).

The interaction between magmatism at Santorini and Kolumbo is still debated. Petrologic studies indicate that their respective crustal magmatic systems are not linked (Klaver et al., 2016), although earthquake tomography suggests a possible connection below 5-km depth (Dimitriadis et al., 2010). The NE-SW strike of regional volcanism, combined with NE-SW striking fault zones, has led other researchers to suggest an interaction between magmatism and tectonic stresses (e.g., Dimitriadis et al., 2009; Feuillet, 2013). Similar NE-SW trending faults are also present at other large Aegean volcanoes (Nomikou & Papanikolaou, 2011; Papazachos & Panagiotopoulos, 1993).

3 Experiment Geometry and Data Acquisition

3.1 Seismic Experiment

In November and December of 2015, a three-dimensional, active-source seismic tomography experiment was conducted in the broader Santorini volcano area of the southern Aegean Sea. The experiment covered an area roughly 120 km × 45 km centered on Santorini, with the goal of imaging the crustal magmatic plumbing system beneath the volcano (Figure 2). Data were collected on 91 ocean bottom seismometers (OBSs) and 65 land seismometers distributed around Santorini's caldera, as well as on Anafi, Christiana, and Anydros islands. The OBSs included 30 Woods Hole Oceanographic Institution instruments (4.5-Hz Geospace three-axis geophone) and 61 Scripps Institute of Oceanography instruments (4.5-Hz Sercel L-28 three-component geophone). Both OBS types also had a High Tech HTI-90-U hydrophone. The land seismometers included 60 Mark 1-Hz geophones from the German Research Center for Geosciences Geophysical Instrument Pool in Potsdam and five CMG-40T and Trillium compact (120 s) seismometers from the Aristotle University of Thessaloniki. All OBS instruments had a 200-Hz sampling rate and all land stations used a 100-Hz sampling rate. There were over 14,300 active-source marine shots using a 6,600-in3 36-component air gun array towed at 12-m depth. Average shot spacing was ~150 m along ENE-WSW oriented shot lines spaced at 1–2-km intervals and included shot-receiver offsets up to ~115 km. Additional azimuthal coverage was achieved using a lower shot-density NW of Santorini and stations located on the neighboring Anafi island (Figure 2a).

3.2 Data Return and Data Quality

The OBS data return was sufficient with one OBS lost and an additional 10 OBSs with noisy and/or poor-quality records on both the hydrophone and vertical channels. A sporadic ~6-Hz ringing noise was observed on several Scripps Institute of Oceanography stations that did not correlate to any external processes such as wind, lightning, and boat traffic. This noise was of variable amplitude and occurred randomly on all four channels. Figure 3 provides an example of good quality OBS data for two stations located in the Christiana and Anydros Basins. Differences between the basins are clearly observed, with large, rather impulsive first arrivals with short coda in the Christiana Basin and more chaotic arrivals with a “ringy” character and longer coda in the Anydros Basin.

Details are in the caption following the image
Seismic data recorded in basins east and west of Santorini. Seismic record sections (hydrophone) for shot lines in the (top) Christiana and the (bottom) Anydros Basins shown with a reduction velocity of 6 km/s. Shots in the Christiana Basin are characterized by a large impulsive first motion in contrast to the more “ringy” character of those in the Anydros Basin. There is good agreement between travel time picks (red) and predicted travel times through our preferred model (blue). Regions shaded grey correspond to near-shot ranges (0–6 km) which were excluded from the inversion (see text for details).

Santorini land stations had variable quality data. Those located on metamorphic rocks and exposed bedrock had great return; in contrast, those situated on volcanic deposits were noisy and difficult to pick. Clear impulsive arrivals were seen on Anafi stations, as all these stations were installed on metamorphic basement, providing good longer-range azimuthal coverage to the experiment. However, the Anafi stations were located outside the velocity model used in this paper, and thus, picks for these stations were not included.

For upper crustal Pg waves crossing the caldera area, arrivals were highly attenuated and travel times for these stations were difficult to pick due to emergent waveforms (Figure S1). For seismic waveforms that did not travel within the footprint of the volcano, clear, impulsive arrivals were observed, indicating that any attenuative structure was located beneath the caldera.

4 Data Processing and Tomographic Inversion

4.1 OBS Relocation

The OBSs were relocated on the seafloor by minimizing the misfit between predicted and observed acoustic water-wave arrival times (Creager & Dorman, 1982) for short-range arrivals (0–2 km; longer ranges were included when necessary). A constant water velocity of 1.52 km/s (determined from expendable bathythermographs) was used and a pick error of 5 ms was estimated. Station locations were fixed to the seafloor using the bathymetric map from Hooft et al. (2017). OBS depths ranged from ~30 to ~890 m, with typical depth uncertainties of 10 m and horizontal uncertainties less than 4 m. Using the precision of the water-wave arrivals (most stations were fit to water-wave travel time RMS values of ~5 ms), we were able to identify and correct several errors in the data set including origin time offsets and mislocated shots. In addition, a nonlinear response at short ranges (typically <0.8 km) was observed on the hydrophone for many instruments, especially shallow stations. This effect did not influence picking of the first motion of water waves, although it did affect the resulting waveform shape (Figure S2). Finally, we were able to identify and correct for a linear drift in the OBS clock for one instrument that had not been accounted for in the initial data processing.

4.2 Pg Travel Time Data

We picked over 200,000 Pg first arrivals using two approaches. Initial picking was conducted using an autopicker from the open-source opendTect software (https://www.dgbes.com/). Several tens of thousands of arrivals with high signal-to-noise ratio traces (e.g., short ranges, predominately less than 20-km range) were collected using this approach. The picking error for these arrivals was assigned a value of 10 ms, following visual inspection of the automatic picks. After this first step, arrivals were picked manually. Errors were visually assigned during the picking process and ranged from 5 ms (high signal-to-noise ratio, exceptional quality data) to 30 ms (low signal-to-noise ratio); the median error was 10 ms with standard deviation of 13 ms. The difference in assigned errors resulted from both variability in the noise level, as well as differences in waveform shape and arrival time due to surface and subsurface complexity. Data were picked on the hydrophone, vertical, and/or a scaled linear stack of the hydrophone and vertical channels. All picks (manual and automatic) were made on the first negative to positive zero crossing (using the polarity convention of the hydrophone channel; Figure 3) as it is the easiest portion of the first arrival pulse to pick. This zero-crossing followed a small negative first motion that was only easily observed in the highest-quality data. All picks were made on waveforms filtered with a fourth-order causal Butterworth filter of 5–25 Hz, for shots with distance ranges between 0 and 65 km. Picking resulted in a high-quality data set for distances between 4 and 30 km at most stations.

4.3 Tomographic Inversion

We inverted the Pg travel time first arrivals using the approach of Toomey et al. (1994), minimizing the squared residual between observed and predicted travel times while also penalizing against both the magnitude and roughness of model perturbations (roughness measured using Laplacian smoothing). The slowness model was defined on a 120 × 45 × 12 km (x, y, and z dimensions) rotated rectangular grid for the forward problem, with a grid spacing of 200 m for both horizontal (x and y) and vertical (z) dimensions, and bathymetry reflected by vertical shearing of the grid (Toomey et al., 1994). The grid was rotated 25.5° counterclockwise from north, to align the x axis with the shot-line orientation (Figure 2a). Travel times through the crust were calculated using the graph theory approach of Moser (1991), with the water-wave segment of the travel time (shot to seafloor) calculated on a grid with a 50 × 50-m horizontal spacing, using a constant water-wave velocity of 1.52 km/s. For the linearized model inversion, a perturbational grid of 400 m × 400 m × 200 m in the horizontal and vertical directions, respectively, was used, inverting for slowness perturbations. Updated slowness values were then linearly interpolated onto the forward problem grid. Iterating on this linearized inversion strategy allowed for an accurate approximation to the nonlinear travel time tomographic problem (see Toomey et al., 1994 for additional details).

A number of inversions were conducted to create a smooth three-dimensional (3-D) VP starting model, so that large-velocity contrasts resulting from the metamorphic horsts and sediment-filled grabens were included in the starting model. To first obtain the best 1-D velocity model for the region (Figure 4), a 3-D inversion was conducted using a 1-D VP starting model derived from a combination of prior gravity, seismic refraction, and receiver function studies (Bohnhoff et al., 2006; blue line in Figure S3). The 3-D output velocity model was averaged at each depth and the new 1-D model was used as an updated starting model. This process was repeated and allowed for the migration to an optimized regional 1-D velocity model (black line; Figures 4 and S3). To obtain the smooth 3-D VP starting model, we used this regional 1-D velocity model and inverted for 3-D isotropic velocity variations, which were then spatially smoothed with a median filter of 5 km by 5 km by 2 km in the horizontal (x and y) and vertical (z) directions, respectively.

Details are in the caption following the image
The 1-D P wave velocity-depth profiles. Averaged 1-D velocity-depth profile of four selected areas around Santorini (outlined in subfigure), compared to the average regional velocity profile (black line). Basin fill is assumed for velocities less than 4 km/s and metamorphic rocks and intrusives for velocities faster than 4 km/s—this corresponds to a steep gradient in the velocity-depth curve. The light blue curve highlights the P wave velocity for the Anydros metamorphic horst (Figure 2), showing faster than average velocities of the metamorphic rocks at shallow depths (0–2 km). At >2-km depth, areas on the western side of Santorini (red and magenta) are systematically faster than those on the eastern side (dark blue and light blue). Background grey scale shows the probability density function distribution of all 1-D velocities of the tomography model.

For the final inversion we used horizontal and vertical smoothing parameters of 200 and 100, respectively, and penalized model perturbations relative to the previous model iteration using a penalty of 1 (see Toomey et al., 1994 for more detail). Each inversion consisted of five model iterations, ensuring the convergence of the RMS travel time misfit (Figure S4). We conducted tens of trial inversions and the main model features presented in this paper were insensitive to reasonable variations of the inversion parameters (Table S1 and Figures S4 and S5). Ranges shorter than 6 km were poorly fit in the initial inversions, a result of the lack of short-range travel time picks and a large grid spacing (200 × 200 × 200 m) relative to the ranges. Picks associated with these ranges were therefore removed from the inversions. The model presented here (Figure 5) was fit to an RMS travel time misfit of 15 ms, which corresponds to a χ2 of 2.2.

Details are in the caption following the image
Velocity anomalies relative to the 1-D average velocity-depth profile at 0-, 1-, 2-, and 3-km depth. Background is illuminated bathymetric map from Hooft et al. (2017). Grey lines are faults defined using the seismic tomography images. These faults are colored black on the depth slice on which they were defined. Dashed black lines are large faults outside the region of resolution. Abbreviations are the same as in Figure 2b.

To ensure the accurate recovery of structures, initial inversions only used the higher-quality, often shorter-range data (<15 km). By predicting expected (theoretical) arrival times through this initial tomographic model, we were able to identify mispicked arrivals and facilitate the manual picking of noisy first arrivals, which were skipped in the initial picking effort. Using this approach, we extended the picked data set from <15 km (high signal-to-noise) to >30 km (often lower signal-to-noise).

5 Results and Interpretation

5.1 Upper Crustal Velocity Variations

The tomographic model shows substantial P wave velocity variations in the upper crust of the study area, with lateral differences in velocity exceeding 3 km/s near the surface (Figure 5). In the topmost kilometer, regions of anomalously high (+1.5 km/s) and low (−1.5 km/s) relative velocity correspond to mapped metamorphic basement and sedimentary basins, respectively (Figures 2, 5a, and 5b). At depths greater than 2 km, the velocity west of Santorini is higher and spatially more uniform compared to the eastern side of the Santorini volcano, where velocities are both lower and more variable (Figures 4 and 5b). At 3-km depth, the overall lateral velocity variability across the study area is still ~2 km/s (±1 km/s; Figure 5d). The reliability of the recovered longer-wavelength lateral velocity variations is validated by checkerboard resolution tests (Text S1 and Figures S6S8), which show that features with length scales of 5 km and greater are well recovered throughout the model, while features with length scales of 3 km are well recovered within the higher-velocity areas, mainly metamorphic basement.

Rapid velocity changes between the low-velocity sedimentary basins and higher-velocity metamorphic basement rocks are interpreted as faults. Near-surface faulting, delineated by sharp velocity changes observed in the 0- and 1-km-depth slices (heavy black lines in Figures 5a and 5b), is in good agreement with faults observed in the topographic-bathymetric map (Figure 2). At the depth of 2 and 3 km, we interpret similar sharp spatial velocity gradients as faults within or beneath the basins (heavy black lines in Figures 5c and 5d).

5.2 Comparison of Tomographic Velocity Model With Seismic Reflection Images

In general, first-order geotectonic features, such as large-scale sedimentary basins and their bounding faults, are well resolved in both tomographic and prior seismic reflection data sets and show similar geometries and structures. Figure 6 shows the location of seven cross sections where detailed model evaluation and interpretation were performed, of which profiles A-A′, B-B′, and C-C′ have associated seismic reflection results (Figure 7).

Details are in the caption following the image
Map showing the location of seismic cross sections. A-A′, B-B′, and C-C′ (purple lines) indicate the locations of the three seismic reflection profiles (from prior studies) and corresponding velocity cross sections (this study) that are shown in Figure 7. D-D′, E-E′, F-F′, and G-G′ (red lines) show the locations of the four interpreted tomography cross sections through regional basins shown in Figure 9. Tomographically defined faults (black lines) are after Figure 5. Orange regions mark areas of diffuse microseismicity (Bohnhoff et al., 2006; Brüstle et al., 2014).
Details are in the caption following the image
Comparisons of tomography velocity model with seismic reflection profiles. Top panels depict seismic reflection profiles from previous studies (locations shown in Figure 6). Middle panels show seismic velocity from this study overlain on the seismic reflection image, after converting depth into a two-way travel time (TWTT). Bottom panel is the same seismic velocity cross section shown as a function of depth. (a) Seismic reflection profile (H14) from Figure 6 of Nomikou et al. (2018) across the Anydros Ridge showing tomographic recovery of the Anafi-Amorgos Basin (AAB) as well as recovery of secondary faulting on the Anydros Horst (AH). (b) E-W seismic reflection profile across the east-central Christiana Basin (Tsampouraki-Kraounaki & Sakellariou, 2018, Figure 7) showing the correlation between the seismic velocity structure and faults in the seismic reflection data. The authors interpret Unit 6 as Messinian evaporites and the region of updoming of this layer correlates with higher seismic velocities. Red lines in bottom panel are faults from the seismic reflection data. (c) S-N seismic reflection profile across the eastern portion of the Christiana Basin (Tsampouraki-Kraounaki & Sakellariou, 2018, Figure 5) showing thin sediments over an old basin margin. Shallow faults from seismic reflection studies (red lines) correlate well with deeper undulations of basement structure.

Figure 7a compares the seismic tomography cross section to a multichannel seismic profile that crosses the Anydros Basin, Anydros Horst, and Anafi-Amorgos Basin from NW to SE (Nomikou et al., 2018; Figure 6). The transition across a large-offset normal fault into the Anafi-Amorgos Basin accurately coincides with an abrupt change from high to low velocities in the tomographic image. A smaller basin bounded by secondary faults is similarly expressed tomographically in the Anydros Horst footwall. The transition from sediment to the underlying basement of the Anafi-Amorgos Basin corresponds approximately to the 4-km/s velocity contour. Additional comparisons with the multichannel seismic results (Nomikou et al., 2016) in the vicinity of Kolumbo submarine volcano are presented in Figures S9 and S10.

A seismic tomography comparison to a W-to-E seismic reflection profile within the Christiana Basin (Tsampouraki-Kraounaki & Sakellariou, 2018) is presented in Figure 7b. In the reflection image, the basin is filled by sediments (Units 1–5; Figure 7b) intercalated with pyroclastic flows (Roman numerals), with Unit 6 proposed to be Messinian evaporites, a late Miocene marker (Tsampouraki-Kraounaki & Sakellariou, 2018). In the center of our tomographic velocity profile, elevated seismic velocities relative to the surrounding sediments (3–4 versus 1.5–3 km/s) correlate with an updoming of the Messinian evaporites, an observation consistent with research showing that evaporites are seismically faster than sediments and pyrolastic flows (e.g., Zong et al., 2017). Small-offset faults within the sediments on the eastern end of the reflection profile correlate with underlying variations in basement structure in the tomography images.

Figure 7c shows a S-N seismic reflection profile through the Christiana Basin (Tsampouraki-Kraounaki & Sakellariou, 2018) and its comparison with tomographic results. Generally, the main basin is seismically slow, whereas the basement rock (Unit 7 in top figure) is seismically fast (~3.5 to 5.5 km/s). In the northern portion of the basin the dip of the margin is well recovered. An old buried basin margin on the southern edge of the Christiana Basin, thought to have become inactive ~1 Ma (Tsampouraki-Kraounaki & Sakellariou, 2018), is also tomographically identified as seismically faster than the main portion of the basin. Near-vertical faults observed in the reflection images (red lines) appear to spatially correlate with variations in the underlying basement, similar to Figure 7b. These faults have been proposed to be strike-slip features associated with a prior episode of deformation, which ceased activity ~1 Ma, as evidenced by the undisturbed sediments overlying the faults (Tsampouraki-Kraounaki & Sakellariou, 2018).

While the travel time tomographic velocity model is less detailed than the seismic reflection images with respect to structure and deposition sequence, it provides valuable velocity and depth constraints, as well as fuller three-dimensional spatial coverage than the existing reflection imaging. The tomographic images also reveal deeper features such as faults that ultimately control the shallower structures observed in the seismic reflection images.

5.3 Mapping Geometry and Thickness of Sedimentary Basins From the Tomography Model

We outline basin depth using a velocity of 4 km/s throughout the model and map both the two-way travel time and the depth to the 4-km/s iso-velocity contour (Figure 8). Visual comparison between seismic reflection images and tomography results indicates that this value typically corresponds to the sediment-basement interface (e.g., Figures 7 and S11). The 4-km/s velocity also corresponds to the midpoint of a rapid increase in average seismic velocity with depth (Figure 4), interpreted as the transition from seismically slow sediments to the underlying, seismically fast metamorphic basement rocks. We contour the depth to the 4-km/s iso-velocity to delineate the 3-D geometry and thickness of the sedimentary basins across the entire study area (Figure 8a), and not just the profiles that were mapped using two-way travel time from seismic reflection studies (e.g., Nomikou et al., 2016).

Details are in the caption following the image
Sedimentary basin thickness maps from the seismic tomography model. (a) Basin thickness, determined using the 4-km/s velocity contour as the sediment-basement transition, with relevant geological features labeled. (b) Tomographic basin thickness converted to TWTT for comparison with seismic reflection studies. The Anydros Basin, despite being a prominent tectonic feature on bathymetric maps, has relatively shallow sediment fill (1–1.5 km). In contrast, the Amorgos-Anafi Basin and the Santorini-Anafi Basins as well as the relatively inactive Christiana Basin contain thick sediment fill (>2 km). Faults after Figure 5, with deeper faults (2–3-km depth) shown in light grey and shallower faults (0–1 km) in dark grey. Dark shading indicates areas with poor resolution of sediment thickness. Labels are the same as in Figure 2.

Figure 8 shows how the faults identified in Figure 5 correlate with, and control, the geometry and internal structure of the basins. Cross sections through the tomography results on two SW-NE cross sections through the Christiana Basin west of Santorini and two WNW-ESE cross sections through the Anydros and Anafi Basins east of Santorini are shown in Figure 9 (with locations of the profiles in Figure 6). This figure validates and elucidates the faults identified in the map views (Figure 5) and their relationship to basin structure (Figure 8).

Details are in the caption following the image
Seismic velocity cross sections through the Christiana and Anydros Basins. (a) Profiles D-D′ and E-E′ through the Christiana Basin in velocity (v) and velocity anomalies (dv) with interpretations of main fault geometries shown in black on dv cross sections. The open triangles mark where the profiles cross the black faults shown on Figure 6. These profiles show evidence for down-dropped blocks within the Christiana Basin. (b) Profiles F-F′ and G-G′ through the Anydros and Anafi Basins, following conventions described above. These two basins show thinner sedimentary fill than the Christiana Basin, more pervasive faults at depth, and are regionally characterized by a seismically slower basement velocities than the profiles through the Christiana Basin (D-D′ and E-E′; see also Figure 4).

5.4 Western Side of Santorini: Christiana Basin and Christiana Ridge

The anomalously low P wave velocities at 0–2-km depth in the western part of the model coincide with the NW-SE striking Christiana Basin (Figures 2 and 5). Low-velocity anomalies associated with the basin extend to ~1-km depth in the basin center and down to 2 km in its SE and NW parts, indicating the presence of two deeper subbasins (Figures 5 and 9). The SE subbasin is deepest close to Santorini, near Christiana island (Figure 5d). Results from previous seismic reflection studies suggest that these subbasins formed in the Early to Middle Pleistocene, before continued expansion was accommodated by wider faults in the north and south (Tsampouraki-Kraounaki & Sakellariou, 2018).

The Christiana Basin is separated from the Sea of Crete to the south by the Christiana ridge (Tsampouraki-Kraounaki & Sakellariou, 2018; Figures 2 and 5). The northern, bathymetrically expressed margin of the ridge proper (dashed black line north of CR in Figures 5a and 5b) has little tomographic expression, showing a minor velocity contrast across the fault (Figure 9a). The tomographically expressed fault margin (solid black line above CR in Figure 5b) is located basinward (north) of the ridge proper, striking NW-SE, and reflecting the presence of an old basin margin, buried under a thin veneer of sediments (Tsampouraki-Kraounaki & Sakellariou, 2018; see also Figure 7c). This fault, which dips at an angle of 25–40°, marks the deepening of the basin to kilometer-scale depths (Tsampouraki-Kraounaki & Sakellariou, 2018; Figures 8 and 9a).

The northern margin of the Christiana Basin is marked by a large SW facing normal fault. While this fault is clearly evident in Figures 8 and 9a, it has little bathymetric expression close to Santorini (Figure 2), most likely because it is covered by volcanic deposits. On the contrary, in the western part of the model away from Santorini, it is identified in both the tomography model and the seafloor bathymetry as seen in the D-D′ and E-E′ model cross sections and their proposed interpretation (Figure 9a). At ~2-km depth, and basinward of the northern basin-bounding fault, there is a large buried normal-fault block, dipping to the SSW at an angle of 25–40° (solid black line north of CB in Figure 5 and solid black line labeled buried block in Figure 9a). This fault forms the northern margin of the SE subbasin.

5.5 Eastern Side of Santorini: Anydros Basin, Anafi Basin, and Anydros Horst

The two large NE-SW trending basins on the eastern side of Santorini, the Anydros, and Anafi Basins are tomographically identified at 0–1-km depths as seismically slow regions (Figures 5a, 5b, and 9b). The Pliocene/Quaternary-aged Anydros Basin has thin basin fill and appears to be only ~1.5 km thick (Figures 8 and 9b). The similarly aged Anafi Basin is divided into two subbasins by a buried ridge at <1-km depth (Figure 5), the Anafi-Amorgos Basin to the north and the Santorini-Anafi Basin to the south. Both subbasins are up to 2 km thick. The buried ridge dividing these subbasins has an ENE-WSW orientation (Figure 5b).

On the eastern side of Santorini, we observe tomographically resolved variations in the strike of basin margins of up to 20° from a mean NE-SW direction. For example, the northern margin of the Anydros Basin, the Ios fault (IF in Figure 5), strikes approximately N40°E, whereas the southern margin of the Anydros Basin strikes roughly N25°E. Similarly, the large dipping fault on the northern margin of the Anafi-Amorgos Basin (the Santorini-Amorgos fault) is curved and has strikes ranging from N25°E to N60°E (Figure 5), a pattern consistent with seafloor morphology (e.g., Hooft et al., 2017; Nomikou et al., 2018; Figure 2b).

The Anydros and Anafi Basins are separated by the Anydros Horst (AH), which has typical bedrock velocities of 5–6 km/s at shallow depths (<1 km), in contrast to the low velocities in the basins (2–4 km/s; Figure 9b). Close to Santorini, this horst is buried under a thin cover of sediments, possibly by volcanic eruption deposits, as suggested by the presence of a relatively flat bathymetry but a strong high-velocity tomographic signature.

5.6 Structures Associated With Santorini

Several NE-SW striking faults extend from the western side of the study area (e.g., Christiana Basin region), through Santorini, to the eastern side (Anydros Horst and Anafi Basin region). At shallow depths, metamorphic blocks that are part of the preexisting volcanic basement and that outcrop in southern Santorini are well recovered as seismically fast regions (Figures 5a and 10). The NE-SW striking Akrotiri fault on the western side of Santorini, which is identified in the bathymetry south of Santorini (Figure 2; Hooft et al., 2017), is tomographically observed at 0–1-km depth as a prominent velocity contrast that strikes through Santorini and out to the east toward the Anydros Horst region (Figures 5a and 5b) juxtaposing the seismically slow basin sediments of the Anydros Basin to the north against the seismically fast metamorphic basement to the south. This fault, which extends from the southwestern termination of the Christiana Basin, through Santorini, to the margin of the Anydros Horst on the east, is subparallel to the regional alignment of volcanic centers (Figures 5a and 5b). At deeper depths (2–3 km), a tomographically defined fault strikes through the northern caldera basin, passing through NE Santorini. This fault may also be connected to a bathymetrically expressed fault extending NE from Christiana island (Figure 2b). From Santorini's northern caldera, the fault extends northeast toward Kolumbo volcano (Figures 5c and S10).

Details are in the caption following the image
Comparison of the seismic velocity beneath Santorini with volcanic features and exposed geology. (a) Geologic map of Santorini (modified from Druitt et al., 1999 and Druitt et al., 1989 with dike locations and orientations (Browning et al., 2015; Fabbro et al., 2013), the Kolumbo and Kameni tectono-magmatic lineaments (defined by prior studies), locations of small-scale faulting (Druitt et al., 1999), location of the 2011–2012 inflation source (Parks et al., 2015), long-wavelength gravity data (black contours at 2-mGal intervals; Budetta et al., 1984), and a LBA (Late Bronze Age) deposit thickness map from seismic reflection data (modified from Johnston et al., 2015). (b) Comparison of previously proposed tectono-magmatic lineament locations (red lines), 2011–2012 inflation point source (blue dot), and Late Bronze Age (LBA) deposit thickness (modified from Johnston et al., 2015) with relative velocity (dv) at 1-km depth. An increase in the LBA deposit thickness directly overlies a decrease in velocity associated with caldera collapse (Hooft et al., 2019) and also overlies the 2011–2012 inflation point source at ~4-km depth (Parks et al., 2015). (c) Comparison of locations of exposed dikes, their orientations, faults, and long-wavelength gravity with dv at 3-km depth. The Kolumbo line (labeled) is clearly observed in the tomography, dividing slower seismic velocities to the SE from faster velocities to the NW. The orientation of the NE-SW striking line parallels exposed dike orientations and matches the strike of regional faults. The Kameni line (labeled) is also observed in the tomography, paralleling the Kolumbo line.

At 1–2-km depth a large circular, low-velocity anomaly is observed in the center of the northern caldera basin, which has been proposed to be related to caldera collapse (Hooft et al., 2019; Figure 10). This anomaly directly overlies the inflation source from a 2011–2012 unrest period (inferred source depth of ~4 km; Parks et al., 2015) and is bounded by the two dominant tectono-magmatic lines on Santorini, the Kameni and Kolumbo lines (Pfeiffer, 2001). The southern margin of the anomaly marks the location of recent volcanism (Kameni lavas; Figure 10; Pyle & Elliott, 2006). Its shape matches that of the shallow basin-fill of about 100 m thick, observed in previous seismic reflection images and inferred to correspond to LBA eruption deposits (Johnston et al., 2015).

6 Discussion

6.1 Differences Between the East and West Sides of Santorini and Relationship to Faulting

The area east of Santorini is seismically active and exhibits ongoing, distributed faulting including a MS = 7.4 normal-faulting event on the Santorini-Amorgos fault in 1956 (Brüstle et al., 2014; Konstantinou, 2010; Okal et al., 2009; Papadopoulos & Pavlides, 1992). There are also two regions of small earthquakes: those around Kolumbo that are thought to result from magmatic processes (Bohnhoff et al., 2006; Dimitriadis et al., 2009) and clustered earthquakes around the AH that may reflect upward migration of fluids within a zone of tectonic weakness or the development of volcanic activity (Bohnhoff et al., 2006; e.g., Figure 6). In contrast, the west side is characterized by the seismically quiet, older Pliocene faulting of the Christiana Basin (e.g., Bohnhoff et al., 2006). These differences have led to the interpretation that the AH is part of a boundary that accommodates the relative motion between a competent Cycladic block and a less-competent eastern Aegean region, dividing the Aegean into a seismically quiet west and more active east (e.g., Bohnhoff et al., 2006; McClusky et al., 2000; Reilinger et al., 2010).

At ~3-km depth the basement east of Santorini has a lower average seismic velocity than to the west of Santorini by ~0.4 km/s (7%; Figures 4 and 5), which can be attributed to the presence of dense, distributed faulting in AH region, reducing rock competency. Preliminary plots of travel time misfit as a function of azimuth show NE-SW oriented seismic anisotropy which results from a fracture network in this region (Figure S12). Focal mechanisms of shallow earthquakes from the AH area (Dimitriadis et al., 2009; Friederich et al., 2014) are also consistent with normal faulting due to NW-SE 130–155° extension, indicating ongoing NE-SW fault activity in this region.

An alternative explanation for the difference in average velocity on the east and west sides of Santorini is that it is generated by differences in composition. While the presence of carbonates, flyschs, greenschists, blueschists, and granites (Kilias et al., 2013) indicates that variable seismic velocities due to compositional differences should exist, the orientation of the geological units is mainly controlled by N-S dipping regional detachments, which is unlikely to produce significant NE-SW velocity variations. We conclude that the association of the lower P wave velocities east of Santorini with active deformation and pervasive fracturing of the basement is the most plausible scenario.

6.2 Tectonic Evolution of the Basins and Christiana-Santorini-Kolumbo Magmatism: Localization of Magma in a Proto-Anydros Basin

Using the delineated faults and basin structures, we refine an existing model for the tectonic evolution of the Santorini region from the Miocene-Pliocene to present (Piper et al., 2007) and investigate how tectonism interacts with, and localizes, magmatism (Figure 11). While the tomographic velocity images provide information on the strike and style of faults, absolute time constraints on fault activity rely on other studies, mainly seismic reflection imaging.

Details are in the caption following the image
Fault evolution and its relationship to magmatism: sketch of the evolution of faulting and volcanism in the Santorini area using the faults defined from the tomographic model (Figure 5) and building on the tectonic evolution model of Piper et al. (2007). Temporal constraints are taken from previous studies (Piper et al., 2007). Red lines and red shading show active faults during each time period, hatched areas indicate basins, red circles indicate regions of volcanic activity, and black arrows indicate approximate extension direction for each period. (a) Formation of the Christiana Basin as a full-graben structure in the Pliocene under NNE-SSW extension. (b) Transition to NW-SE extension leading to Anydros Horst formation from the Pliocene-Pleistocene on, with initial formation of a proto-Anydros Basin. (c) Initiation of Christiana volcanism in the Pleistocene. (d) Volcanism at Santorini and Kolumbo developed in the Quaternary within the proto-Anydros Basin. Current diffuse seismicity (orange region; Bohnhoff et al., 2006; Brüstle et al., 2014) occurs predominately on the eastern side of Santorini.

The earliest stage of faulting in the region, during the Miocene-Pliocene (Figure 11a), resulted in WNW-ESE oriented faults that define the Christiana Basin and likely the buried ENE-WSW faults that divide the Anafi Basin (Anastasakis & Piper, 2005; Piper et al., 2007; hereafter these faults will be referred to as E-W to be consistent with the literature). The offset between these two sets of tomographically resolved faults (Figures 5 and 11a) supports the inference of Piper et al. (2007) that a transfer zone may have existed striking approximately north through the area where Santorini currently resides.

In the late Pliocene to Pleistocene (Figure 11b), these predominately E-W structures were crosscut by NE-SW striking faults (Piper et al., 2007). This may have occurred as a result of NW-SE extension due to differences in motion between the Cycladic and eastern Aegean blocks (e.g., Reilinger et al., 2010). These faults formed primarily on the eastern side of Santorini (Figures 5 and 11b) and the earlier E-W fault systems became mostly inactive (e.g., Piper et al., 2007). The tomographic results indicate that a proto-Anydros Basin formed by progressive normal faulting along parallel to subparallel NE-SW trending faults near the current Anydros Basin and extending further to the SW through Santorini toward Christiana island (Figure 11b). A substantial portion of the proto-basin is currently filled by the Santorini volcanic edifice (bold black line; Figure 4c) and the Kolumbo volcanic field (Figures 11b and 11d). Gravity and magnetic data support the presence of a NE-SW oriented fault zone underlying the caldera (Figure 10; Budetta et al., 1984). We infer that the proto-Anydros Basin intersected the Christiana Basin in the region of the SE Christiana subbasin, potentially contributing to deepening of the SE portion of this subbasin (Figure 8).

During the Pleistocene (Figure 11c), volcanism initiated at Christiana Island (~1.2 Myr; Piper et al., 2007) and was roughly localized at the intersection of the proto-Anydros Basin with the Christiana Basin (Figure 11c). Because the proto-Anydro Basin is filled by the Santorini and Kolumbo edifices (Figure 11d), the NW-SE extension that formed this basin must have preceded the majority of these volcanic eruptions. Volcanism at Santorini began around the Middle Pleistocene (650 ka; Figure 11c; Druitt et al., 1999). Seismic reflection data show that the present-day Anydros Basin opened in six tectonic pulses (Hübscher et al., 2015; Nomikou et al., 2016), creating the basin that is currently observed to the NE of Santorini. The exact relationship between Santorini volcanism and Anydros Basin formation is not known, although the seismic reflection data indicate that eruptions from nearby Kolumbo seamount started after Anydros Basin initiation (Nomikou et al., 2018). Because all of the regional volcanism falls within the proto-Anydros Basin (Figure 11d), it is likely that this basin has played a role in localizing volcanism. Such interaction between extension and volcanism is readily observed in analog models where volcanism is localized in rift basins during oblique extension (Corti et al., 2003).

The above interpretations, where proto-Anydros Basin formation precedes major local volcanism, are consistent with previous models that inferred that the rotation from older NNE-SSW extension to the current NW-SE extension in the early Pliocene was associated with an increase in volcanism in the area (Piper et al., 2007). Across the Hellenic arc, volcanism is dominantly associated with NE-SW striking features (i.e., NW-SE extension; e.g., Budetta et al., 1984) and only secondarily with NW-SE strikes (Kokkalas & Aydin, 2013).

6.3 Volcano-Tectonic Lineaments

At Santorini we recover and improve on the geometry of the well-known Kameni and Kolumbo volcano-tectonic lineaments, using information inferred from the seismic tomography images (Figures 5, 8, and 10). The northern margin of the low-velocity anomaly within the Santorini caldera is bordered by the Kolumbo line, initially identified as an alignment of volcanic edifices in the northern portion of Santorini. It has been suggested that one of the vents of the LBA eruption was located along this line (Pfeiffer, 2001). In our study, the Kolumbo line is tomographically imaged at 3-km depth as a linear feature that divides a seismically fast region to the north from a lower velocity region to the south, oriented more northerly than suggested by previous studies and subparallel to the regional NE-SW trending faults (Figures 2 and 10). The tomographically defined Kolumbo lineament is associated with faulting observed in multichannel seismic profiles north of Kolumbo volcano (Figure S10; Nomikou et al., 2016). This line is also aligned with exposed dikes in the northern portion of the caldera walls and with a similarly oriented dike exposed on Therasia on the western side of the caldera (Browning et al., 2015; Fabbro et al., 2013). The joint interpretation of these observations suggests that (at depth) the Kolumbo line is controlled by regional tectonism and that it plays a dominant role in localizing magmatic and volcanic features, both within, and close to, Santorini.

In contrast to the Kolumbo line, the Kameni line has a weaker tomographic signature. This line, which is thought to have focused vents of prior explosive eruptions (Druitt et al., 1989), has previously been defined using the orientation of post-LBA eruption vents on the Kameni islands (Pyle & Elliott, 2006). Moreover, it was re-activated seismically in the recent 2011–2012 seismo-volcanic crisis (e.g., Newman et al., 2012), verifying its active character. Our study shows that the line is associated with a NE-SW trending region of limited lateral extent that divides anomalously slow velocities to the NW from faster velocities to the SE (Figure 10) and lies parallel to the tomographically recovered Kolumbo line.

The Kameni and Kolumbo lines border the caldera low-velocity anomaly, a region of collapse in the northern caldera (Hooft et al., 2019). This isolated region of deepened caldera collapse, which overlies a recent influx of magma (e.g., Newman et al., 2012; Parks et al., 2015), seems to have been limited in extent by these tectono-magmatic lineaments. Similarly, the majority of exposed dikes and vents associated with Santorini volcanism also fall between the Kolumbo and Kameni lines. It is clear these lineaments have helped to control the tectonic evolution of the volcanic system and, given their association with eruptive vents, they have likely shaped magma input into the upper crust.

7 Conclusions

We used the seismic velocity model to outline faults and basins (Figure 5) and to correlate magmatic and tectonic processes. In the Santorini region, we find that the tomographic results compare well with previous seismic reflection studies and geological outcroppings on Santorini (Figure 10). Geologically observed dikes and volcanic chains around the volcano directly overlie tomographically observed faults and tectono-magmatic lineations. Inside the caldera, volcanic vents and deposits from the Late Bronze Age eruption correspond to a seismically slow region of caldera collapse. Our results and interpretations, combined with those from other studies, support the following conclusions:
  1. Resolved tomographic structures, specifically orientations of faults and basins, are consistent with a previously proposed tectonic evolution model consisting of NNE-SSW extension during the Miocene-Pliocene transitioning to NW-SE extension in the Late Pliocene-Pleistocene that predominates on the eastern side of Santorini (Figure 11). This present-day extension is associated with pervasive faulting east of Santorini.
  2. The Christiana, Santorini, and Kolumbo volcanic centers are found to lie in a NE-SW trending basin-like structure, the proto-Anydros Basin (Figure 5). These results are in agreement with models that suggest volcanism initiated after the transition to NW-SE extension in the Pleistocene.
  3. Two tectono-magmatic lineaments control magma emplacement at Santorini. These lineaments border a NE-SW trending region of low velocity, are associated with dikes in the caldera walls, and strike parallel to active NE-SW trending regional faults and subparallel to regional volcanic chains. The two lineaments also bound a region of localized caldera collapse within Santorini's northern caldera, verifying the strong relationship between tectonics and magmatism.


We thank the R/V Langseth crew, the marine science technicians, and the office of marine management, as well as the OBSIP personnel for the help in the execution of this project. We also thank the personnel involved in the land station deployment and recovery. We thank undergraduates Kathleen Walls and Janelle Jacobson for their help with the picking the initial data set. Data used in this research were provided by instruments from the Ocean Bottom Seismograph Instrument Pool (http://www.obsip.org) which is funded by the National Science Foundation. OBSIP data are archived at the IRIS Data Management Center (http://www.iris.edu). The Geophysical Instrument Pool Potsdam provided 60 land seismometers. The Aristotle University of Thessaloniki contributed five land seismometers and the Greek military donated helicopter time for installations on the smaller islands. This work benefited from access to the University of Oregon high-performance computer, Talapas. The experiment and analysis were supported by the National Science Foundation under grant OCE-1459794 to the University of Oregon and Leverhulme Trust grant RPG-2015-363 to Imperial College London. The 3-D velocity model is available in the Marine Geoscience Data System with data doi:10.1594/IDEA/324827 and on GeomapApp (http://www.geomapapp.org). Several of the figures were made using Generic Mapping Tools (https://github.com/GenericMappingTools/gmt).