The 2018 Mw 6.8 Zakynthos, Greece,Earthquake: Dominant Strike-Slip Faultingnear Subducting SlabEfthimios Sokos*1, František Gallovič2, Christos P. Evangelidis3, Anna Serpetsidaki1, Vladimír Plicka2,Jan Kostelecký4, and Jiří Zahradník2

Abstract

Cite this article as Sokos, E., F. Gallovič,C. P. Evangelidis, A. Serpetsidaki, V. Plicka,J. Kostelecký, and J. Zahradník (2020). The2018 Mw 6.8 Zakynthos, Greece,Earthquake: Dominant Strike-Slip Faultingnear Subducting Slab, Seismol. Res. Lett.XX, 1–12, doi: 10.1785/0220190169.

Supplemental Material

With different styles of faulting, the eastern Ionian Sea is an ideal natural laboratory toinvestigate interactions between adjacent faults during strong earthquakes. The 2018Mw 6.8 Zakynthos earthquake, well recorded by broadband and strong-motion net-works, provides an opportunity to resolve such faulting complexity. Here, we focuson waveform inversion and backprojection of strong-motion data, partly checked bycoseismic Global Navigation Satellite System data. We show that the region is undersubhorizontal southwest–northeast compression, enabling mixed thrust faulting andstrike-slip (SS) faulting. The 2018 mainshock consisted of two fault segments: a low-dip thrust, and a dominant, moderate-dip, right-lateral SS, both in the crust. Slip vectors,oriented to southwest, are consistent with platemotion. The sequence can be explainedin terms of trench-orthogonal fractures in the subducting plate and reactivated faults inthe upper plate. The 2018 event, and anMw 6.6 event of 1997, occurred near three local-ized swarms of 2016 and 2017. Future numerical models of the slab deformation andocean-bottom seismometer observations may illuminate possible relations amongearthquakes, swarms, and fluid paths in the region.

IntroductionMultiple faults acting during an earthquake have been generallywell known, but on global scale less observations have beenavailable for near-simultaneous ruptures of different faultingmechanisms, specifically in subduction zones. For example,Lay et al. (2013) reported a doublet of twoMw ∼ 7 events belowJapan trench, where a thrust faulting (TF) along the subductioninterface was followed after 14 s by a shallower normal faultingin the overriding plate. A rare evidence of a thrust event on aplate interface, which triggered a normal-faulting event in theoverriding plate was provided for an Mw ∼ 7 earthquake in theChile subduction zone (Hicks and Rietbrock, 2015). Particularlychallenging in terms of strain partitioning are the regions wheresubduction terminates, and plate motions continue along trans-form faults. A good example is the 2016 Mw 7.8 earthquake inNew Zealand, in which a dominant strike-slip (SS) faultingoccurred in the upper plate and possibly triggered minor slip onthe underlying subduction thrust (Mouslopoulou et al., 2019;Ulrich et al., 2019). Resolving fault complexity for strong earth-quakes (Mw 6.0–6.9) in the shallowest parts of the subductiontermination zones is even more challenging, and this articlefocuses on such a task in western Greece.

The major ongoing convergent tectonic process in westernGreece is subduction, imaged by seismic tomography and

other structural studies (Spakman et al., 1993; Laigle et al.,2004; Suckale et al., 2009; Sachpazi et al., 2016; Halpaap et al.,2019). The active overriding of the Aegean plate over the sub-ducting African plate, derived from Global Navigation SatelliteSystem (GNSS) data, is oriented toward southwest, approxi-mately perpendicular to the trench, being consistent with slipvectors of many earthquakes (Kiratzi and Louvari, 2003;Hollenstein et al., 2006; Shaw and Jackson, 2010).

Zakynthos (or Zante) Island is situated at a subduction-ter-mination zone, a part of the Ionian Islands–Akarnania Block(IAB) (Pérouse et al., 2017). At the northwest edge, this block isseparated from Apulian–Ionian microplate by the right-lateralCephalonia transform fault. The southwest boundary of IAB isthe Hellenic subduction backstop front. The northeast boun-dary of IAB is a mixture of SS and extensional structures (e.g.,the Corinth Gulf). The southeast boundary of IAB has not beenwell known until the 2008Mw 6.3 Movri Mountain earthquake

1. Department of Geology, Seismological Laboratory, University of Patras, Patras,Greece; 2. Faculty of Mathematics and Physics, Charles University, Prague, CzechRepublic; 3. National Observatory of Athens, Institute of Geodynamics, Athens,Greece; 4. Faculty of Mining and Geology, University of Ostrava, Ostrava, CzechRepublic

*Corresponding author: esokos@upatras.gr

© Seismological Society of America

Volume XX • Number XX • – 2020 • www.srl-online.org Seismological Research Letters 1

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020

(Gallovič et al., 2009; Konstantinou et al., 2009; Serpetsidakiet al., 2014), which proved activity of a blind right-lateral trans-form fault crossing the western Peloponnese and possibly con-tinuing further toward the southwest into the Ionian Sea. Theinterior of IAB, south of Zakynthos, is a zone of “diffuse defor-mation” (Pérouse et al., 2017). Although being continuouslyseismically active, and being obviously related to shallow sub-duction process, its faulting style has not been understood yet.In this respect, the recent 2018 Mw 6.8 event plays an impor-tant role for seismotectonic interpretations in western Greece.Lessons learned here may apply also for other subduction-ter-mination zones (Mouslopoulou et al., 2019), where the defor-mation partitioning between the slab and upper plate may takea variety of forms.

Knowledge of the crustal structure of the studied region hasbeen significantly improved by a mixed onshore and offshore(ocean-bottom) temporary seismic network (Papoulia et al.,2014). The latter study illuminated the spatial variation ofthe Moho depth, and provided a layered velocity modelconsistently used throughout this article. A weakly northeast-dipping (∼5°) seismic reflector has been detected at the depthof ∼10–15 km, supposedly mapping the top of the subductingplate in the area west and southwest of Zakynthos (Clémentet al., 2000; Laigle et al., 2004), situated ∼10 km aboveMoho (Halpaap et al., 2019). Lacking more detailed informa-tion, in the following parts of this article we use the well-mapped Moho depth of Papoulia et al. (2014) and plot the slabtop 10 km above the Moho. It provides an approximate (sche-matic) location of the slab top, at the depths 10–20 km, close tothe above-cited reflector depths.

In the instrumental era, the region near Zakynthos experi-enced three Mw 6–7 earthquakes, roughly every 20 yr (1959,1976, 1997, and 2018), consistently with a high seismic cou-pling (Laigle et al., 2002; Chousianitis et al., 2015). No histori-calM > 7 event has been documented. On 25 October 2018, at22:54 UTC, an Mw 6.8 earthquake occurred southwest ofZakynthos. It caused limited damage on the island and nocasualties (Institute of Engineering Seismology and EarthquakeEngineering [ITSAK], 2018). The event was observed globally,and its broad characteristics were soon outlined as follows.The Global Centroid Moment Tensor (Global CMT) projectsuggested a centroid depth of 12 km, scalar momentM0 � 2:3 × 1019 N · m, and strike/dip/rake angles of13°/24°/165°, an oblique-TF mechanism. The Global CMT sol-ution comprised a significant non-double-couple (non-DC)component, namely a compressional compensated linearvector dipole, CLVD � −44%. The European data centersreported centroid depths <20 km, with focal mechanismsranging from the mixed SS and thrust type to an almost pureSS, often with a notable non-DC component. For example,the National Observatory of Athens (NOA) published themoment tensor (MT) with CLVD of −61%. Interestingly, anMw 4.8 foreshock was a pure low-dip thrusting mechanism

(strike/dip/rake ∼300°=10°=100°), similar to four majorMw 5+ aftershocks; however, smaller aftershocks were of bothtypes, thrust and SS. Our preliminary analysis, reported toEuropean Mediterranean Seismological Centre two weeks afterthe event, speculated about a segmented fault (Zahradníket al., 2018).

The goal of this article is to improve understanding of thecomplex faulting style taking place near Zakynthos, comple-menting our previous earthquake studies of the Ionian Seaislands of Lefkada and Cephalonia (Sokos et al., 2015, 2016).To this goal, we analyze source process of the 2018 mainshockand aftershocks using regional broadband, accelerometric, andGNSS data, considering also the 2011–2018 seismicity of theregion. We interpret the mainshock in terms of a segmentedsource model, possibly related to trench-orthogonal fracturesin the subducting plate and reactivated faults in the upper plate.

Source ModelingPoint-source models of mainshock andaftershocksThe mainshock nucleated ∼45 km southwest of Zakynthos,close to a local bathymetric low (the sea depth of 4 km; seeFig. 1). Exact hypocenter position is unknown because a smallforeshock of an unknown position and magnitude precededthe mainshock by a few seconds, thus complicating the arrival-time picking. For the same reason, the first-motion polaritiesare problematic. We made a probabilistic location (Lomaxet al., 2001), see Text S1, and Figures S1 and S2 (available inthe supplemental material to this article), pointing to a shallowdepth, and hereafter we use the epicenter (latitude/longitude37.27°/20.43°) corresponding to the arbitrarily fixed sourcedepth of 5 km, with origin time of 22:54:47.5 UTC. None ofthe following modeling methodologies relies on the particularhypocenter position.

A significant Mw 4.8 foreshock occurred 32 min before themainshock (Fig. S1). The mainshock was followed by a stan-dard exponentially decaying aftershock sequence that we firstlocated with Hypoinverse (Klein, 2002), and then relocatedwith hypoDD code (Waldhauser, 2001). Their median formalerrors are ∼1 km, and a few hundred meters, respectively. Thesequence included one Mw 5.1 event 15 min after mainshock,and four other events of Mw 5+ in the first week. After ∼80days, another, smaller sequence of Mw < 4 appeared (seeFig. 1b), situated basically in the same region as the mainshock.In the first day after the mainshock, the activity occupied anarea of ∼20 × 20 km, lacking any clear spatial pattern. Later, amuch broader area (∼60 × 60 km) was activated; at least threeaftershock streaks of the southwest–northeast orientationappeared near Zakynthos, gradually intensified with time, andanother similarly oriented streak was created south of themainshock since the beginning of the sequence (Fig. 1). Thispattern is stable with respect to the used velocity models.However, relative to our reference model (Papoulia et al., 2014),

2 Seismological Research Letters www.srl-online.org • Volume XX • Number XX • – 2020

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020

a systematic mean vertical shift of about 3 km downward and5 km upward occurs in the models of Papadimitriou et al.(2012) and Haslinger et al. (1999), respectively. The activatedvolume has the 10–20 km depth range. Considering the loca-tion inaccuracies and the limited knowledge of the slab top,the events occurred at or near the slab top, but definitely abovethe Moho determined by Papoulia et al. (2014). The aftershockstreaks were rather vertical fault structures in the upperplate (Fig. 2).

The MT of the mainshock, particularly the strike/dip/rakeangles (hereafter, s/d/r) are much better determined than thehypocenter because we calculate these parameters by invert-ing relatively low-frequency seismic waveforms (Sokos andZahradník, 2013; Zahradník and Sokos, 2018). Using broad-band stations in Greece and Italy at distances 270–630 kmand frequencies 0.01–0.02 Hz, our centroid appeared at lati-tude/longitude = 37.39°/20.63°; it was an event of the “odd”(mixed) type (Frohlich, 1992), with a large negative CLVDcomponent (s/d/r = 12°/41°/165°, M0 � 1:7 × 1019 N · m,DC ∼ 40%, CLVD ∼ −60%; see Fig. 1 and Table S1).

MTs of the strongest aftershocks were also calculated byfull-waveform inversion (Table S2). We obtained a varietyof focal mechanisms, ranging from TF, to SS, and mixed types.Their spatial distribution indicates that in the western part ofthe sequence we observe mostly TF, whereas in the streaks wehave rather SS. The focal mechanisms provide an estimate ofstress field (Vavryčuk, 2014). The σ1 axis of the maximumcompression is well resolved featuring plunge ∼20° and azi-muth of ∼230° (see Fig. 1b). This azimuth is close to theplate-motion vectors (Hollenstein et al., 2008). The eigenvaluescorresponding to the stress axes σ2 and σ3 have similar mag-nitudes, the shape ratio is ∼0:75, and hence the orientation of

these two axes is rather ambiguous, explaining the coexistenceof the TF and SS events in the region.

Before constructing a finite-fault model, it is useful tounderstand the shear (pure-DC) part of the rupturing process.To this goal, we made multiple-point source (MPS) modelingin a DC-constrained mode (Zahradník et al., 2005, 2017;Zahradník and Gallovič, 2010; Zahradník and Sokos, 2014;Sokos et al., 2016). Full waveforms of 11 near-regionalstrong-motion accelerometer stations were inverted, using avariety of trial source positions, velocity models, and frequencyranges (e.g., Figs. S3 and S4). They robustly indicate that thelargest DC point-source contribution (major subevent), situ-ated near the centroid, has a moderately dipping SS mecha-nism, s=d=r ∼ 10°=40°=180°, hereafter simply called SS.Although the other significant point-source subevents wereless stable in terms of their position and mechanism, many fea-tured a low-dip TF type, for example, s=d=r ∼ 300°=10°=60°,acting almost simultaneously with the SS subevent(within ∼� 5 s).

−4000

−4000

−200

0

−200

0

0 10 20

km

20° 20.2° 20.4° 20.6° 20.8° 21° 21.2°

0

10

20

30

40

50

60

70

80

90

100

110

Num

ber

of d

ays

0

10

20

30

Dep

th (

km)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17 18

19

20

2122

23

24

25 26

27

1

3

2

22˚

38˚

Mediterranean ridge

Aegean Sea

Greece

CTF

SB

S

20° 20.2° 20.4° 20.6° 20.8° 21° 21.2°

°

°

°

°

°

°

° ° ° ° ° ° °

° ° ° ° ° ° °

Strofades

26.2

mm

/yr

(a) (b)

Figure 1. The 2018 Zakynthos earthquake sequence.(a) Foreshock Mw 4.8 (yellow star and focal mechanism plot),mainshock epicenter (green star), centroid moment tensor (CMT;green square and focal mechanism plot), and activity in the first24 hr, superimposed on bathymetry. The Cephalonia transformfault (CTF) and the subduction backstop (SBS) are shown. Threecharacteristic plate-velocity vectors are included (Hollensteinet al., 2008). Inset demonstrates a broader area; tectonic linesafter Faccenna et al. (2014). (b) Activity in the first 120 days withfocal mechanisms (see Table S2); note the southwest–northeast(SW-NE) streaks, numbers 1–4. Principal stress axes are shown inthe inset.

Volume XX • Number XX • – 2020 • www.srl-online.org Seismological Research Letters 3

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020

The idea of the mainshocksource process involving theTF and SS contributions isplausible for several reasons:(1) the TF mechanism is anobvious candidate for a largeevent in subduction environ-ment. (2) Both TF and SSmechanisms were observed inthe aftershock sequence.(3) The TF + SS combinationcould explain the large negativeCLVD of the mainshock.(4) Despite the different faulttypes, the indicated TF andSS subevents share similar ori-entation of slip vectors, if bothrupture their low-dip nodalplanes. Then, the latter indi-cates that the moderately dip-ping SS fault is right lateral.

Segmented fault modelof the mainshockWe used the linear slip inver-sion (LSI) method (Gallovičet al., 2015; Gallovič, 2016;Mai et al., 2016; Pizzi et al.,2017), in which the ruptureprocess is discretized in spaceand time along the assumedfault planes. Model parametersare the spatial–temporal slip-rate samples, spanning thewhole rupture duration. Theinversion result is thus pre-sented in terms of full slip-ratefunctions instead of derivedkinematic rupture parameterssuch as rise time, rupturespeed, and so on. The sameaccelerometer stations as inthe MPS modeling were used(see Fig. S5). Importantly, thehypocenter position is not usedin the LSI method, and thus itsonly rough knowledge, for theZakynthos earthquake, doesnot affect the performance ofthe inversion. Nevertheless,slip inversion is uneasy, pos-sibly containing artifacts,because the epicentral

(b)

0

10

20

30

Dep

th(k

m)

0 10 20 30 40 50 60 70 80 90

NW1 SE1

±14 km

0

10

20

30

Dep

th(k

m)

0 10 20 30 40 50 60 70 80 90

NW2 SE2

±10 km

0

10

20

30

Dep

th(k

m)

0 10 20 30 40 50 60 70 80 90

Distance (km)

NW3 SE3

±12 km

0

10

20

30

Dep

th (

km)

0 10 20 30 40 50 60 70

SW1 NE1

±12 km

0

10

20

30

Dep

th (

km)

0 10 20 30 40 50 60 70

SW2 NE2

±12 km

0

10

20

30

Dep

th (

km)

0 10 20 30 40 50 60 70 80

SW3 NE3

±7.5 km

0

10

20

30

Dep

th (

km)

0 10 20 30 40 50 60 70 80

Distance (km)

SW4 NE4

±10 km

(a)

20°

20°

20.2°

20.2°

20.4°

20.4°

20.6°

20.6°

20.8°

20.8°

21°

21°

21.2°

21.2°

37° 37°

37.2° 37.2°

37.4° 37.4°

37.6° 37.6°

37.8° 37.8°

38° 38°

0 10 20

km

Strofades

0

10

20

30

40

50

60

70

80

90

100

110

Num

ber

of d

ays

NW1

SE1

NW2

SE2

NW3

SE3

SW1

NE1

SW2

NE2

SW3

NE3

SW4

NE4

Figure 2. Aftershocks. (a) Activity in the 120 days after mainshock and map view of the northwest–southeast (NW-SE) and SW-NE profiles. Thick black lines refer to major tectonic structures (as inFig. 1). (b) Vertical cross sections along the NW-SE (left) and SW-NE (right) profiles with eventssituated a few kilometers off the profiles, as indicated by the ± symbols. The surface topography(black), schematic slab top (blue dashed line), and the Moho (red line) are also shown. Note thattoward Zakynthos Island the activity is gradually shifted from the slab top to the upper plate, bestseen in the NW3-SE3 section. The same plot also shows that the NW-SE streaks observed in themap view close to Zakynthos are associated with subvertical structures.

4 Seismological Research Letters www.srl-online.org • Volume XX • Number XX • – 2020

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020

distances are relatively large (∼50–180 km) and the azimuthalcoverage is poor (Gallovič and Zahradník, 2011). The inversionneeds to be regularized by smoothing, ensuring slip-rate pos-itivity and prescribing total seismic moment.

In preliminary runs, we considered just a single SS fault,whose position and orientation were suggested by the majorpoint-source subevent. In such a case, some seismic stationswere poorly fitted. Then, we tested several combined SS + TFmodels, as suggested by the MPS modeling. Finally, we proposea source model with two fault segments, providing a very goodfit of waveforms, overall consistency with the distribution andfocal mechanisms of aftershocks, and explaining the observednon-DC part of the overall single-point-source MT.

The proposed model, consisting of two segments (TF andSS), is described in Table S3 and shown in Figure 3. It fits

seismic data in the 0.02–0.15 Hz range with variancereduction (VR) of 0.66 (seeFig. 4 and Fig. S5), better thanthe single-fault model (VR =0.52, see Fig. S6). The space–time evolution of the slip rate(see Fig. S7) can be character-ized as follows. The momentrelease started on the TF seg-ment near the shallow hypo-center and propagated tonortheast along both faults.The major slip-rate episodeswere identified; two of themoccurred on the TF segment(∼6–9 s and 16–19 s after ori-gin time, centered at the depthsof 6 and 16 km, respectively),and one appeared on the SSsegment (∼9–13 s, centered atthe depth of 13 km). The SSfault ruptured a single compactpatch, with peak slip of 0.8 m,situated between the two TFepisodes. The entire thrustprocess developed on or nearthe slab top, but due to limiteddata and location inaccuracieswe cannot be more specific.The dominant SS segment,likely situated in the upperplate, cannot be easily associ-ated with any aftershock clus-ter. The ∼6 s delay betweenthe observed shallow nuclea-tion and the deeper significantslip can be attributed to weak

nucleation as observed, for example, in dynamic inversionof the 2016 Amatrice, Italy, earthquake (Gallovič et al.,2019). We note that these main aspects of the source modelare preserved even when removing station LTHK from theinversion as it is the closest one, supposedly affecting the inver-sion significantly (Fig. S8).

Importantly, if summing the DC-constrained MTs of the TFand SS faults (shown in Fig. 3c), we obtain a significant CLVDvalue (−56%), thus explaining the observed non-DCmechanismof our single-point MT solution (CLVD ∼ −60%), in agreementwith other agencies, for example, Global CMT and NOA(Fig. S2), reporting CLVD of −44% and −61%, respectively.Obviously, even more complex models cannot be ruled out.

Although we believe that major features of the presentedsource models are realistic, some details cannot be taken at face

0.0

0.2

0.4

0.6

0.8

Slip

m

0

5

10

15

20

Tim

e

s

(a)

(b)

0

0.5

1

1.5

2

2.5

0 2 4 6 8 10 12 14

Mom

ent r

ate

(×10

18N

m/s

)

Time (s)

TFSS

CMT inversion

(c)

Slip inversion

20.5°

20.5°

21°

21°

37°

37.5°

38°

0 10 20 30km

TF

Sub 1

Sub 2

20.5°

20.5°

21°

21°

37° 37°

37.5° 37.5°

38° 38°

0 10 20 30km

SS

Sub 1

Sub 2

Figure 3. Segmented fault model. (a) Map view of the slip distribution on the modest thrust (left)and dominant strike-slip (SS; right) faults is shown by color. Overprinted are aftershocks (dots) andspace–time high-frequency energy release from backprojection (small circles, shade-codedaccording to time; time 0 corresponds to origin time). Star marks the epicenter; rectangles cor-respond to the respective fault segments, and their top edges are marked by thick lines. Slip vectorsare shown by arrows at the black focal mechanism plots. Subevents of the two-point model (blue,denoted Sub 1 and Sub 2), taken from Figure S3, are included for comparison. More details are inFigures S5–S11. (b) Moment-rate time functions of the two fault segments. (c) Comparison of theCMT (left) and the MT obtained by summing the double-couple tensors of the thrust faulting (TF)and SS segments from the slip inversion (right).

Volume XX • Number XX • – 2020 • www.srl-online.org Seismological Research Letters 5

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020

value due to limited azimuthal coverage. For example, we can-not strictly rule out that the source models as a whole areshifted by ∼10 km in the northwest–southeast direction, asindicated by the location uncertainty (Fig. S1b). Also, somelinear source features (e.g., the aftershock streaks or slip-distribution patches) could be somewhat lengthened in thenorthwest–southeast direction (see Gallovič and Zahradník,2011). The latter warning is particularly relevant for the back-projection, which—due to involvement of high frequencies—might be less robust than slip models.

To further validate the source model, we considerablyincreased the frequency band (2–8 Hz) and performed back-projection of strong-motion records at local distances, basicallyof the Sg wave group (Evangelidis and Kao, 2014; Evangelidis,2015). For methodical details, see Text S2 and Figure S9. Asshown in Figure 3, we observe a well-focused backprojectionpattern, systematically moving northeast away from epicenterwith increasing time. These independent data agree very wellwith the finite-fault inversion, confirming the dominant space–time progression of the source process in the southwest–north-east direction.

Finally, we considered also coseismic static GNSS data (seeText S3 and Table S4; Kostelecký and Karský, 1987; Koubaand Héroux, 2001). The nearest permanent stations (ZAKY andSTRF) were situated at similar epicentral distances (∼44 km) onZakynthos and Strofades islands. Both recorded ∼5 cm horizon-tal displacements, pointing toward south-southwest and south-southeast, respectively. An ∼2 cm motion toward southwest wasrecorded on station PYRG. Our seismic source model agrees withZAKY and PYRG (Fig. S5). For STRF, we fit the displacement

orientation but underestimate the amplitude by a factor of ∼4.The GNSS amplitude fit at STRF could be remarkably improvedin a joint inversion of the seismic and GNSS data; such anapproach adds a significant slip patch near STRF station(Fig. S10), similar as if we invert only GNSS data (Fig. S11).Nevertheless, we consider this patch as rather unrealistic forthe following reasons. There are no indications for such a slippatch from the backprojections (Fig. 3). There are also no earlyaftershocks close to this patch. Duration of this patch is about15 s, starting as soon as 2 s after the origin time (although being∼40 km far from the hypocenter). The consequence of such longduration is that this patch does not effectively generate any seis-mic signal in the frequency range considered. Therefore, we ten-tatively associate the observed STRF amplitude with a triggeredrapid aseismic movement, perhaps similar to that proposed forprevious earthquakes by Stiros (2005) to explain unusually largedisplacements on the Strofades island. For example, very largemotion (∼12 cm) was recorded at STRF during the Mw 6.6earthquake of 1997 (Hollenstein et al., 2006).

DiscussionAlthough the low-dip TF part of the source process is naturallyacceptable in the shallow part of the subduction zone, the

3.89AAMF

1.77FSK

1.72GUR

2.06SERG

8.55LTHK

3.43PDO

4.55LCHA

1.78LXRA

2.61PATC

2.70MTHA

3.04ZARA

N-S E-W Z Displ.(cm)

100 s

20°

20°

20.5°

20.5°

21°

21°

21.5°

21.5°

22°

22°

22.5°

22.5°

37° 37°

37.5° 37.5°

38° 38°

38.5° 38.5°

0 50km

AAMF

FSK

GUR

SERG

LTHK

PDO

LCHA

LXRAPATC

MTHA

ZARA

(a) (b)

Figure 4. Waveform fit for the two-segment fault model. (a) Mapof stations used in waveform inversion (with epicenter, faultplanes and the 2008Movri Mountain earthquake shown in blue).(b) Real (black) and synthetic (red) displacements are shown inthe 0.02–0.15 Hz frequency range, VR = 0.66. Peak observeddisplacements are shown to the right.

6 Seismological Research Letters www.srl-online.org • Volume XX • Number XX • – 2020

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020

dominant SS fault slip of the 2018 earthquake is a significantfinding deserving more discussion (Fig. 5). A frequentlyevoked explanation of SS faulting near the trench as transpres-sional strain partitioning due to oblique plate convergence(Wesnousky, 2005) is not likely in our case because (1) our

region does not feature aclearly oblique plate motionrelative to the trench, and(2) such SS faults should belargely trench-parallel, withstrike N45°W, while our SSfault is striking at N10°E.Nevertheless, shear strain androtation, possibly generatingSS faulting, have been docu-mented near Zakynthos byGNSS measurements of platevelocities (Hollenstein et al.,2008; Chousianitis et al.,2015). Moreover, shearingstrain was proposed to explainthe right-lateral SS faulting ona vertical buried fault inwestern Peloponnese in the2008 Mw 6.3 MovriMountain earthquake (Fig. S5;Serpetsidaki et al., 2014). Theoffshore continuation of theMovri fault was also hypoth-esized (Pérouse et al., 2017).Strain is likely accommodatedby trench-orthogonal (south-west–northeast) fracturescreated in response to segmen-tation of the subducting slaband piecewise rollback(Sachpazi et al., 2016).

Instead of the transpres-sional strain partitioning, wepropose two alternative inter-pretations: our SS segmentstriking at ∼10° could be gen-erally explained as a fault off-set, for example, a restrainingbend or a part of a SS duplex,in our case possibly connectingone of the aftershock streaksnear Zakynthos (streak num-ber 3 in Fig. 1) with thesouthern streak (number 4).These streaks represent verticalfaults in the upper plate, and,interestingly, they have a

geometrical similarity with the major trench-orthogonal,right-lateral fault system in the slab (Sachpazi et al., 2016).

Another possible interpretation, independent of the trench-orthogonal fractures and the aftershock streaks, is that our10°-striking right-lateral SS occurred on an approximately

(a)

(b)

Figure 5. Faulting geometry: An overview. Slip on the thrust and SS segments are presentedtogether with aftershocks (see legend). Two selected snapshots are shown here; (a) view to NEfrom an elevation of 50°, and (b) roughly the same view direction from an elevation of 20°.Snapshots are taken from the MATLAB code provided as supplemental material.

Volume XX • Number XX • – 2020 • www.srl-online.org Seismological Research Letters 7

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020

north–south-oriented reverse fault or normal fault that wasreactivated as SS fault (Naliboff et al., 2013) and bounds a relayramp (Rotevatn and Peacock, 2018). Relay zones are highlyfractured, so this interpretation seems to be supported bythe dispersed aftershocks. The moderate 40° dip of our SS faultseems quite acceptable in regions where previous thrust faultsor normal faults were reactivated as SS to accommodate thepresent-day strain (Avouac et al., 2014; Sato et al., 2015).All the mentioned fault types exist in the region (Stiros, 2005;

Kokkalas et al., 2013; Papouliaet al., 2014; Avramidis et al.,2017). We also note that somefeatures of the mainshock, thatis, the shallow nucleation, thedominant SS motion (situatedlikely in the upper plate), andthe minor TF on (or near to)the plate interface resembleother top-to-bottom breakingsystems of subduction termi-nations of the world (e.g.,New Zealand; Mouslopoulouet al., 2019).

In a more general context,we attempt to interpret howlarge events in the region areconnected to long-term, ∼8 yrlasting deformation processes(Fig. 6). The earthquakesequence that occurred nearZakynthos in April 2006 com-prised six events ofMw 5.0–5.7(Global CMT; Serpetsidakiet al., 2010; Papadimitriouet al., 2012). Their totalmoment is Mw 6.7, thus the2006 sequence is a long-termsegmented analogy of theshort-term segmented 2018mainshock. Similar to 2018,the 2006 sequence “activateda single sub-horizontal faultzone at a depth of about13 km, corresponding to theinter-plate subduction boun-dary” (Serpetsidaki et al.,2010). Contrary to the 2018event, all the 2006 earthquakeswere remarkably similar interms of their pure TF mecha-nisms, strike 345°–21°, dip10°–39°, rake 118°–128°(Global CMT), not including

any SS process. Another notable event to be discussed hereis the 1997 Mw 6.6 earthquake that occurred within∼20 km from the 2018 event. Considering the centroid inac-curacies, we can assume that they occurred at the same place.Focal mechanism (s/d/r = 8°/31°/162°, Global CMT) is a mixedTF-SS type with negative CLVD (−27%). Two (TF) subevents,9 s delayed, were indicated by teleseismic inversions (Kiratziand Louvari, 2003). Thus, the 2018 event seems to be a (larger)analogy of the 1997 event.

0

1

2

3

4

5

6

Mag

nitu

de

100 200 300 100 200 300

2016 2017100 200 300 100 200 300

20° 20.2° 20.4° 20.6° 20.8° 21° 21.2°

37°

37.2°

37.4°

37.6°

37.8°

38°

20° 20.4° 20.8° 21.2°

37.2°

37.6°

38°

19972006

2018

0 10 20 30 40

Depth (km)

10 km

Magnitude2345

(a)

(b)

Figure 6. Seismicity 2011–2018. (a) Space distribution of earthquakes 2011–2018 (until the 2018mainshock). The centroids of significant earthquakes southwest of Zakynthos, in 1997 (Kiratzi andLouvari, 2003), 2006 (Serpetsidaki et al., 2010), and 2018 (this article), occurred near two localizedclusters. (b) Magnitude–time distribution in 2016–2017, demonstrating the swarm character of theclusters near the 1997 and 2018 Mw > 6 earthquakes. More details are in Figure S12.

8 Seismological Research Letters www.srl-online.org • Volume XX • Number XX • – 2020

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020

The area is characteristic for its noteworthy long-term back-ground seismicity at depths <40 km. Figure 6 and Figure S12show our relocation (Klein, 2002) based on phase data of theNational Observatory of Athens for the period 2011–2018. InFigure 6, clustering is obvious: (1) one cluster occurred in 2011and 2015 at the place of the 2006 Zakynthos sequence, and(2) one cluster was repeatedly activated three times (twicein 2016 and once in 2017) very near the 1997 and 2018 events.A closer look at the latter cluster in Figure 6b shows that itsactivity was of swarm type (lacking any magnitude-dominantevent). Similar swarms recur in time (irregularly), as shown inbroader time window of Figure S12. To explain the spatial cor-relation of the swarms with the significant earthquakes, we canspeculate that the region is affected by fluids of deep origin.

The localized “chimney” of small events detected down to adepth of >80 km, and reported southeast of Zakynthos, out ofthe region covered by Figure 6 (Papoulia et al., 2014), couldtrack a possible fluid channel from the upper mantle, assuggested in previous studies (Konstantinou et al., 2011;Giannopoulos et al., 2012). Our shallower events rather suggestthat fluids could be channeled up-dip in the slab (Halpaap et al.,2019), and this flow is enhanced by the heavy fracturing ofreactivated relay ramps here (Rotevatn and Peacock, 2018).Alternatively, faults existing in the seismically coupled brittlecrust can be loaded by the partially and episodically creeping ofthe topmost serpentinized part of the upper mantle; then creepin the mantle could activate earthquake swarms, some of whichmay represent foreshocks of large earthquakes (Kuna et al.,2019). It is matter of future study to distinguish betweenthese explanations by geodynamic modeling (Čížková andBina, 2013).

ConclusionFault interactions, specifically in subduction zones, belong tokey problems of earthquake mechanics and seismic hazardassessment. Following previous rare studies of Lay et al.(2013), Hicks and Rietbrock (2015), and Mouslopoulou et al.(2019), we aimed at resolving fault complexity of a strongMw 6.8 earthquake in the shallowest part of the subduction-termination zone in western Greece.

The studied region near Zakynthos is under subhorizontalcompression of azimuth ∼230°, and both thrust (TF) and SSevents are possible. The Mw 6.8 event in 2018 was of a mixedTF + SS type, as was also the whole aftershock sequence. Themainshock consisted of two fault segments: a TF (dip ∼10°)and a right-lateral SS (dip ∼40° and strike ∼10°). Both slip vec-tors have the same southwest-orientation, consistent with theGNSS-derived major plate motion. The two-segment faultmodel explains the observed non-DC radiation. The low dipof the TF segment (similar to many aftershocks) suggeststhe interplate origin. Relatively distant aftershocks (seeminglyoccurring off the mainshock fault) mapped several southwest–northeast streaks, or faults. As a whole, the mainshock

progressed also in the southwest–northeast direction. Thisdirection closely follows the trench-orthogonal fracturesrelated to differential slab motions. The 10°-striking SS seg-ment of the mainshock could be a fault bend connecting suchoffshore structures, or it could be explained by reactivation ofapproximately north–south-trending fault relay zone in theupper plate. Although the 2018 sequence involved a thrustfault, the SS faulting was dominating the mainshock; this resultshould be considered in future seismic and tsunami hazardassessment in the region. The observations could also applyto other subduction-termination zones worldwide.

The previous earthquake of similar focal mechanism (1997Mw 6.6) occurred almost at the same place as the 2018 event.This place, coinciding with a large bathymetric deep, hostedthree earthquake swarms in 2016 and 2017. Precise investiga-tion of the swarms would deserve future deployment of anocean-bottom seismometer network; it could confirm or ruleout possible triggering of the major M 6+ earthquakes in thearea by deep-seated processes such as fluid flows or the loadingeffect of the episodically creeping upper mantle. Future obser-vations should be complemented by numerical modeling of theslab deformation, especially in the shallow depths of the sub-duction-termination zone.

Data and ResourcesWaveform data were sourced from the corresponding Observatoriesand Research Facilities for European Seismology (ORFEUS) EuropeanIntegrated Data Archive (EIDA) node (http://www.orfeus-eu.org/eida)operated by the National Observatory of Athens (http://eida.gein.noa.gr). We used data from the HL (Institute of Geodynamics,National Observatory of Athens, doi: 10.7914/SN/HL), HP(University of Patras, doi: 10.7914/SN/HP), which is cooperating cer-tain stations jointly with the Charles University in Prague, HT(Aristotle University of Thessaloniki, doi: 10.7914/SN/HT), HA(National and Kapodistrian University of Athens, doi: 10.7914/SN/HA), and HI (Institute of Engineering Seismology and EarthquakeEngineering, doi: 10.7914/SN/HI) networks. Global NavigationSatellite System (GNSS) data for STRF station were provided byIstituto Nazionale di Geofisica e Vulcanologia (INGV), Rete IntegrataNazionale Global Positioning System (RING) Working Group (2016),(doi: 10.13127/RING). PYRG and ZAKY GNSS station data were pro-vided by METRICA S.A. The bathymetry maps were based onEuropean Marine Observation and Data Network Digital TerrainModel (EMODNET DTM) (1/16 arcmin grid resolution) obtainedand modified from the Hellenic Centre for Marine Research (filewas provided by D. Sakellariou). Figures were created using GenericMapping Tools (GMT) v.5.4.5 (Wessel et al., 2013). Software ISOLA(developed by J. Z. and E. S.) can be downloaded from https://github.com/esokos/isola and with recent updates from http://geo.mff.cuni.cz/~jz/for_Cuba2018/. All websites were last accessed May2019. The supplemental material for this article includes four tablesand 12 figures, illustrating details of seismicity, focal mechanisms,GNSS data, multiple-point source modeling, and slip inversions. Inaddition, it includes a zip file with text files and a MATLAB scriptenabling 3D view of aftershocks and slip on two fault planes.

Volume XX • Number XX • – 2020 • www.srl-online.org Seismological Research Letters 9

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020

AcknowledgmentsThe authors thank Maria Sachpazi, Sotiris Kokkalas, and Craig Bina foruseful discussions. F. Gallovič and J. Zahradník were supported fromthe Czech Science Foundation Grant Number GACR-18-06716J, andV. Plicka from CzechGeo/EPOS LM2015079. J. Kostelecký acknowl-edges support from Grant Number LO 1506 (Ministry of Educationof Czech Republic). E. Sokos, C. P. Evangelidis, and A. Serpetsidakiacknowledge financial support by the HELPOS Project, “HellenicPlate Observing System” (MIS 5002697). Useful comments ofEditor-in-Chief Allison Bent, Reviewer Stephen P. Hicks, and twoanonymous reviewers greatly improved the article.

ReferencesAvouac, J. P., F. Ayoub, S. Wei, J. P. Ampuero, L. Meng, S. Leprince, R.

Jolivet, Z. Duputel, and D. Helmberger (2014). The 2013, Mw 7.7Balochistan earthquake, energetic strike-slip reactivation of athrust fault, Earth Planet. Sci. Lett. 391, 128–134, doi: 10.1016/j.epsl.2014.01.036.

Avramidis, P., G. Iliopoulos, K. Nikolaou, N. Kontopoulos, A.Koutsodendris, and G. J. van Wijngaarden (2017). Holocene sed-imentology and coastal geomorphology of Zakynthos Island,Ionian Sea: A history of a divided Mediterranean island,Palaeogeogr. Palaeoclimatol. Palaeoecol. 487, 340–354, doi:10.1016/J.PALAEO.2017.09.018.

Chousianitis, K., A. Ganas, and C. P. Evangelidis (2015). Strain androtation rate patterns of mainland Greece from continuous GPSdata and comparison between seismic and geodetic moment release,J. Geophys. Res. 120, no. 5, 3909–3931, doi: 10.1002/2014JB011762.

Čížková, H., and C. R. Bina (2013). Effects of mantle and subduction-interface rheologies on slab stagnation and trench rollback, EarthPlanet. Sci. Lett. 379, 95–103, doi: 10.1016/j.epsl.2013.08.011.

Clément, C., A. Hirn, P. Charvis, M. Sachpazi, and F. Marnelis (2000).Seismic structure and the active Hellenic subduction in the Ionianislands, Tectonophysics 329, nos. 1/4, 141–156, doi: 10.1016/S0040-1951(00)00193-1.

Evangelidis, C. P. (2015). Imaging supershear rupture for the 2014Mw 6.9 Northern Aegean earthquake by backprojection of strongmotion waveforms, Geophys. Res. Lett. 42, no. 2, 307–315, doi:10.1002/2014GL062513.

Evangelidis, C. P., and H. Kao (2014). High-frequency source imagingof the 2011 October 23 Van (Eastern Turkey) earthquake by back-projection of strong motion waveforms, Geophys. J. Int. 196, no. 2,1060–1072, doi: 10.1093/gji/ggt437.

Faccenna, C., F. A. Capitanio, F. Funiciello, F. Horvàth, L. Jolivet, andC. Piromallo (2014). Mantle dynamics in the Mediterranean, Rev.Geophys. 52, no. 3, 283–332, doi: 10.1002/2013RG000444.

Frohlich, C. (1992). Triangle diagrams: Ternary graphs to display sim-ilarity and diversity of earthquake focal mechanisms, Phys. EarthPlanet. In. 75, nos. 1/3, 193–198, doi: 10.1016/0031-9201(92)90130-N.

Gallovič, F. (2016). Modeling velocity recordings of theMw 6.0 SouthNapa, California, earthquake: Unilateral event with weak high-fre-quency directivity, Seismol. Res. Lett. 87, no. 1, 2–14, doi: 10.1785/0220150042.

Gallovič, F., and J. Zahradník (2011). Toward understanding slipinversion uncertainty and artifacts: 2. Singular value analysis, J.Geophys. Res. 116, no. B02309, doi: 10.1029/2010JB007814.

Gallovič, F., W. Imperatori, and M. Mai (2015). Effects of three-dimen-sional crustal structure and smoothing constraint on earthquakeslip inversions: Case study of the Mw6.3 2009 L'Aquila earthquake,J. Geophys. Res. 120, 428–449, doi: 10.1002/2014JB011650.

Gallovič, F., Ľ. Valentová, J.-P. Ampuero, and A.-A. Gabriel (2019).Bayesian dynamic finite-fault inversion: 2. Application to the 2016Mw6.2 Amatrice, Italy, earthquake, J. Geophys. Res. 124, no. 7,6970–6988.

Gallovič, F., J. Zahradník, D. Křížová, V. Plicka, E. Sokos, A.Serpetsidaki, and G. A. Tselentis (2009). From earthquake centroidto spatial-temporal rupture evolution: Mw 6.3 Movri Mountainearthquake, June 8, 2008, Greece, Geophys. Res. Lett. 36, no. 21,doi: 10.1029/2009GL040283.

Giannopoulos, D., E. Sokos, K. I. Konstantinou, A. Lois, and G. A.Tselentis (2012). Temporal variation of shear-wave splittingparameters before and after the 2008 Movri Mountain earthquakein northwest Peloponnese (Greece), Ann. Geophys. 55, no. 5,1027–1038, doi: 10.4401/ag-5586.

Halpaap, F., S. Rondenay, A. Perrin, S. Goes, L. Ottemöller, H.Austrheim, R. Shaw, and T. Eeken (2019). Earthquakes track sub-duction fluids from slab source to mantle wedge sink, Sci. Adv. 5,no. 4, doi: 10.1126/sciadv.aav7369.

Haslinger, F., E. Kissling, J. Ansorge, D. Hatzfeld, E. Papadimitriou, V.Karakostas, K. Makropoulos, H. G. Kahle, and Y. Peter (1999). 3Dcrustal structure from local earthquake tomography around theGulf of Arta (Ionian region, NW Greece), Tectonophysics 304,no. 3, 201–218, doi: 10.1016/S0040-1951(98)00298-4.

Hicks, S. P., and A. Rietbrock (2015). Seismic slip on an upper-platenormal fault during a large subduction megathrust rupture,NatureGeosci. 8, no. 12, 955–960, doi: 10.1038/ngeo2585.

Hollenstein, Ch., A. Geiger, H. G. Kahle, and G. Veis (2006). CGPStime-series and trajectories of crustal motion along the WestHellenic Arc, Geophys. J. Int. 164, no. 1, 182–191, doi: 10.1111/j.1365-246X.2005.02804.x.

Hollenstein, Ch., M. D. Müller, A. Geiger, and H. G. Kahle (2008).Crustal motion and deformation in Greece from a decade ofGPS measurements, 1993-2003, Tectonophysics 449, nos. 1/4,17–40, doi: 10.1016/j.tecto.2007.12.006.

Institute of Engineering Seismology and Earthquake Engineering(ITSAK) (2018). The 26/10/2018 M6.8, south Ionian sea earth-quake, Analysis of Accelerometer Records and Damages, retrievedfrom http://www.itsak.gr/uploads/news/earthquake_reports/EQ_ZAKYNTHOS_20181026_M6.8_v2_el.pdf (last accessedMarch 2019) (in Greek).

Kiratzi, A., and E. Louvari (2003). Focal mechanisms of shallow earth-quakes in the Aegean Sea and the surrounding lands determinedby waveform modelling: A new database, J. Geodyn. 36, nos. 1/2,251–274, doi: 10.1016/S0264-3707(03)00050-4.

Klein, F. W. (2002). User’s guide to HYPOINVERSE-2000: A Fortranprogram to solve for earthquake locations and magnitudes, U.S.Geol. Surv. Open-File Rept. 02-171, available at http://geopubs.wr.usgs.gov/open-file/of02-171/ (last accessed April 2019).

Kokkalas, S., E. Kamberis, P. Xypolias, S. Sotiropoulos, and I.Koukouvelas (2013). Coexistence of thin-and thick-skinned tec-tonics in Zakynthos area (Western Greece): Insights from seismicsections and regional seismicity, Tectonophysics 597/598, 73–84,doi: 10.1016/j.tecto.2012.08.004.

10 Seismological Research Letters www.srl-online.org • Volume XX • Number XX • – 2020

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020

Konstantinou, K. I., C. P. Evangelidis, and N. S. Melis (2011). The 8 June2008 Mw 6.4 earthquake in northwest Peloponnese, western Greece:A case of fault reactivation in an overpressured lower crust? Bull.Seismol. Soc. Am. 101, no. 1, 438–445, doi: 10.1785/0120100074.

Konstantinou, K. I., N. S. Melis, S. J. Lee, C. P. Evangelidis, and K.Boukouras (2009). Rupture process and aftershocks relocation ofthe 8 June 2008 Mw 6.4 earthquake in northwest Peloponnese,western Greece, Bull. Seismol. Soc. Am. 99, no. 6, 3374–3389,doi: 10.1785/0120080301.

Kostelecký, J., and G. Karský (1987). Analysis of 1970-1983 circum-zenithal measurements, Bull. Astron. Inst. Czechosl. 38, 16–23.

Kouba, J., and P. Héroux (2001). Precise point positioning using IGSorbit and clock products, GPS Solut. 5, no. 2, 12–28, doi: 10.1007/PL00012883.

Kuna, V. M., J. L. Nábělek, and J. Braunmiller (2019). Mode of slip andcrust-mantle interaction at oceanic transform faults, NatureGeosci. 12, no. 2, 138–142, doi: 10.1038/s41561-018-0287-1.

Laigle, M., A. Hirn, M. Sachpazi, and C. Clément (2002). Seismic cou-pling and structure of the Hellenic subduction zone in the IonianIslands region, Earth Planet. Sci. Lett. 200, nos. 3/4, 243–253, doi:10.1016/S0012-821X(02)00654-4.

Laigle, M., M. Sachpazi, and A. Hirn (2004). Variation of seismic cou-pling with slab detachment and upper plate structure along thewestern Hellenic subduction zone, Tectonophysics 391, nos. 1/4,85–95, doi: 10.1016/j.tecto.2004.07.009.

Lay, T., Z. Duputel, L. Ye, and H. Kanamori (2013). The December 7,2012 Japan Trench intraplate doublet (Mw 7.2, 7.1) and inter-actions between near-trench intraplate thrust and normalfaulting, Phys. Earth Planet. In. 220, 73–78, doi: 10.1016/j.pepi.2013.04.009.

Lomax, A., A. Zollo, P. Capuano, and J. Virieux (2001). Precise, abso-lute earthquake location under Somma-Vesuvius volcano using anew three-dimensional velocity model, Geophys. J. Int. 146, no. 2,313–331, doi: 10.1046/j.0956-540X.2001.01444.x.

Mai, P. M., D. Schorlemmer, M. Page, J. Ampuero, K. Asano, M.Causse, S. Custodio, W. Fan, G. Festa, M. Galis, et al. (2016). Theearthquake-Source Inversion Validation (SIV) project, Seismol.Res. Lett. 87, no. 3, 690–708, doi: 10.1785/0220150231.

Mouslopoulou, V., V. Saltogianni, A. Nicol, O. Oncken, J. Begg, A.Babeyko, S. Cesca, and M. Moreno (2019). Breaking a subduc-tion-termination from top to bottom: The large 2016 Kaikōuraearthquake, New Zealand, Earth Planet. Sci. Lett. 506, 221–230,doi: 10.1016/j.epsl.2018.10.020.

Naliboff, J. B., M. I. Billen, T. Gerya, and J. Saunders (2013). Dynamicsof outer-rise faulting in oceanic-continental subduction systems,Geochem. Geophys. Geosys. 14, no. 7, 2310–2327, doi: 10.1002/ggge.20155.

Papadimitriou, P., K. Chousianitis, A. Agalos, A. Moshou, E. Lagios,and K. Makropoulos (2012). The spatially extended 2006 AprilZakynthos (Ionian Islands, Greece) seismic sequence and evidencefor stress transfer, Geophys. J. Int. 190, no. 2, 1025–1040, doi:10.1111/j.1365-246X.2012.05444.x.

Papoulia, J., J. Makris, and A. Tsambas (2014). Microseismicity andcrustal deformation of the Kyparissiakos Gulf, south-westernHellenic Arc, using an “amphibious” seismic array and a 3D veloc-ity model obtained from active seismic observations, Boll. Geofis.Teor. Appl. 55, 281–302, doi: 10.4430/bgta0086.

Pérouse, E., M. Sébrier, R. Braucher, N. Chamot-Rooke, D. Bourlès, P.Briole, D. Sorel, D. Dimitrov, and S. Arsenikos (2017). Transitionfrom collision to subduction in Western Greece: The Katouna-Stamna active fault system and regional kinematics, Int. J.Earth Sci. 106, no. 3, 967–989, doi: 10.1007/s00531-016-1345-9.

Pizzi, A., A. Di Domenica, F. Gallovič, L. Luzi, and R. Puglia (2017).Fault segmentation as constraint to the occurrence of the mainshocks of the 2016 Central Italy seismic sequence, Tectonics 36,no. 11, 2370–2387, doi: 10.1002/2017TC004652.

Rotevatn, A., and D. C. P. Peacock (2018). Strike-slip reactivation ofsegmented normal faults: Implications for basin structure and fluidflow, Basin Res. 30, no. 6, 1264–1279, doi: 10.1111/bre.12303.

Sachpazi, M., M. Laigle, M. Charalampakis, J. Diaz, E. Kissling, A.Gesret, A. Becel, E. Flueh, P. Miles, and A. Hirn (2016). SegmentedHellenic slab rollback driving Aegean deformation and seismicity,Geophys. Res. Lett. 43, 651–658, doi: 10.1002/2015gl066818.

Sato, H., N. Kato, S. Abe, A. Van Horne, and T. Takeda (2015).Reactivation of an old plate interface as a strike-slip fault in aslip-partitioned system: Median Tectonic Line, SW Japan,Tectonophysics 644, 58–67, doi: 10.1016/j.tecto.2014.12.020.

Serpetsidaki, A., P. Elias, M. Ilieva, P. Bernard, P. Briole, A.Deschamps, S. Lambotte, H. Lyon-Caen, E. Sokos, and G. A.Tselentis (2014). New constraints from seismology and geodesyon the Mw = 6.4 2008 Movri (Greece) earthquake: Evidence fora growing strike-slip fault system, Geophys. J. Int. 198, no. 3,1373–1386, doi: 10.1093/gji/ggu212.

Serpetsidaki, A., E. Sokos, G. A. Tselentis, and J. Zahradnik (2010).Seismic sequence near Zakynthos Island, Greece, April 2006:Identification of the activated fault plane, Tectonophysics 480,nos. 1/4, 23–32, doi: 10.1016/j.tecto.2009.09.024.

Shaw, B., and J. Jackson (2010). Earthquake mechanisms and activetectonics of the Hellenic subduction zone, Geophys. J. Int. 181,no. 2, 966–984, doi: 10.1111/j.1365-246X.2010.04551.x.

Sokos, E., and J. Zahradník (2013). Evaluating centroid-moment-ten-sor uncertainty in the new version of ISOLA software, Seismol. Res.Lett. 84, no. 4, doi: 10.1785/0220130002.

Sokos, E., A. Kiratzi, F. Gallovič, J. Zahradník, A. Serpetsidaki, V.Plicka, J. Janský, J. Kostelecký, and G.-A. Tselentis (2015).Rupture process of the 2014 Cephalonia, Greece, earthquake dou-blet (Mw6) as inferred from regional and local seismic data,Tectonophysics 656, 131–141, doi: 10.1016/j.tecto.2015.06.013.

Sokos, E., J. Zahradník, F. Gallovič, A. Serpetsidaki, V. Plicka, and A.Kiratzi (2016). Asperity break after 12 years: The Mw 6.4 2015Lefkada (Greece) earthquake, Geophys. Res. Lett. 43, no. 12,6137–6145, doi: 10.1002/2016GL069427.

Spakman, W., S. van der Lee, and R. van der Hilst (1993). Travel-timetomography of the European-Mediterranean mantle down to1400 km, Phys. Earth Planet. In. 79, nos. 1/2, 3–74, doi:10.1016/0031-9201(93)90142-V.

Stiros, S. C. (2005). Geodetic evidence for mobilization of evaporitesduring the 1997 Strophades (W Hellanic Arc) Mw 6.5 earthquake,J. Geophys. Eng. 2, no. 2, 111–117, doi: 10.1088/1742-2132/2/2/005.

Suckale, J., S. Rondenay, M. Sachpazi, M. Charalampakis, A. Hosa,and L. H. Royden (2009). High-resolution seismic imaging ofthe western Hellenic subduction zone using teleseismic scatteredwaves, Geophys. J. Int. 178, no. 2, 775–791, doi: 10.1111/j.1365-246X.2009.04170.x.

Volume XX • Number XX • – 2020 • www.srl-online.org Seismological Research Letters 11

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020

Ulrich, T., A.-A. Gabriel, J.-P. Ampuero, and W. Xu (2019). Dynamicviability of the 2016 Mw 7.8 Kaikōura earthquake cascade on weakcrustal faults, Nat. Comm. 10, no. 1, 1213, doi: 10.1038/s41467-019-09125-w.

Vavryčuk, V. (2014). Iterative joint inversion for stress and faultorientations from focal mechanisms, Geophys. J. Int. 199, no. 1,69–77, doi: 10.1093/gji/ggu224.

Waldhauser, F. (2001). hypoDD: A computer program to computedouble-difference earthquake locations, U.S. Geol. Surv. Open-File Rept. 01-113.

Wesnousky, S. G. (2005). The San Andreas and Walker Lane faultsystems, western North America: Transpression, transtension,cumulative slip and the structural evolution of a major transformplate boundary, J. Struct. Geol. 27, no. 8, 1505–1512, doi: 10.1016/J.JSG.2005.01.015.

Wessel, P., W. H. F. Smith, R. Scharroo, J. Luis, and F. Wobbe (2013).Generic Mapping Tools: Improved version released, Eos Trans.AGU 94, no. 45, 409–410, doi: 10.1002/2013EO450001.

Zahradník, J., and F. Gallovič (2010). Toward understanding slipinversion uncertainty and artifacts, J. Geophys. Res. 115,no. B09310, doi: 10.1029/2010JB007414.

Zahradník, J., and E. Sokos (2014). The Mw 7.1 Van, EasternTurkey, earthquake 2011: Two-point source modelling by iterative

deconvolution and non-negative least squares, Geophys. J. Int. 196,no. 1, 522–538, doi: 10.1093/gji/ggt386.

Zahradník, J., and E. Sokos (2018). ISOLA code for multiple-pointsource modeling—Review, in Moment Tensor Solutions: AUseful Tool for Seismotectonics, S. D'Amico (Editor), SpringerInternational Publishing, Cham, Switzerland, 1–28, doi:10.1007/978-3-319-77359-9_1.

Zahradník, J., H. Čížková, C. R. Bina, E. Sokos, J. Janský, H. Tavera,and J. Carvalho (2017). A recent deep earthquake doublet in lightof long-term evolution of Nazca subduction, Sci. Rep. 7, ArticleNumber 45153, doi: 10.1038/srep45153.

Zahradník, J., A. Serpetsidaki, E. Sokos, and G. A. Tselentis (2005).Iterative deconvolution of regional waveforms and a double-eventinterpretation of the 2003 Lefkada earthquake, Greece, Bull.Seismol. Soc. Am. 95, no. 1, 159–172, doi: 10.1785/0120040035.

Zahradník, J., E. Sokos, and V. Plicka (2018). Zakynthos 25/10/2018,Mw 6.8 earthquake: Superposition of strike-slip and thrust? retrievedfrom https://www.emsc-csem.org/Files/news/Earthquakes_reports/Zakynthos_2018_Report.pdf (last accessed April 2019).

Manuscript received 6 July 2019

Published online 15 January 2020

12 Seismological Research Letters www.srl-online.org • Volume XX • Number XX • – 2020

Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220190169/4921268/srl-2019169.1.pdfby Charles University useron 16 January 2020