Investigation of the March 2021 Strong Earthquake Sequence in Thessaly, Central Greece

Home , Domeniko, Epicenter, Morphotectonics, Tyrnavos

geosciences

Article Seismic and Geodetic Imaging (DInSAR) Investigation of the March 2021 Strong Earthquake Sequence in Thessaly, Central Greece

Gerassimos A. Papadopoulos 1,*, Apostolos Agalos 1 , Andreas Karavias 2, Ioanna Triantafyllou 3, Issaak Parcharidis 2 and Efthymios Lekkas 3

1 International Society for the Prevention & Mitigation of Natural Hazards, 10681 Athens, Greece; [email protected] 2 Department of Geography, Harokopio University, 17671 Athens, Greece; [email protected] (A.K.); [email protected] (I.P.) 3 Department of Dynamic Tectonic Applied Geology, Faculty of Geology and Geoenvironment, National and Kapodistrian University of Athens, 15784 Athens, Greece; [email protected] (I.T.); [email protected] (E.L.) * Correspondence: [email protected]

Abstract: Three strong earthquakes ruptured the northwest Thessaly area, Central Greece, on the 3, 4 and 12 March 2021. Since the area did not rupture by strong earthquakes in the instrumental period of seismicity, it is of great interest to understand the seismotectonics and source properties Citation: Papadopoulos, G.A.; of these earthquakes. We combined relocated hypocenters, inversions of teleseismic P-waveforms Agalos, A.; Karavias, A.; and of InSAR data, and moment tensor solutions to produce three fault models. The first shock Triantafyllou, I.; Parcharidis, I.; ◦ ◦ (Mw = 6.3) occurred in a fault segment of strike 314 and dip NE41 . It caused surface subsidence Lekkas, E. Seismic and Geodetic −40 cm and seismic slip 1.2–1.5 m at depth ~10 km. The second earthquake (Mw = 6.2) occurred to Imaging (DInSAR) Investigation of the NW on an antithetic subparallel fault segment (strike 123◦, dip SW44◦). Seismic slip of 1.2 m the March 2021 Strong Earthquake − Sequence in Thessaly, Central Greece. occurred at depth of ~7 km, while surface subsidence 10 cm was determined. Possibly the same ◦ ◦ Geosciences 2021, 11, 311. https:// fault was ruptured further to the NW on 12 March (Mw = 5.7, strike 112 , dip SSW42 ) that caused doi.org/10.3390/geosciences11080311 ground subsidence −5 cm and seismic slip of 1.0 m at depth ~10 km. We concluded that three blind, unknown and unmapped so far normal fault segments were activated, the entire system of which Academic Editors: forms a graben-like structure in the area of northwest Thessaly. Ioannis Koukouvelas, Riccardo Caputo, Tejpal Singh Keywords: March 2021 earthquakes; Thessaly area; Central Greece; hypocenter relocation; and Jesus Martinez-Frias teleseismic P-waveforms inversion; InSAR data inversion; Okada displacement; blind normal faults; antithetic faults; graben-like structure Received: 7 June 2021 Accepted: 21 July 2021 Published: 25 July 2021 1. Introduction Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in During March 2021 the area of northwest Thessaly, Central Greece, was ruptured published maps and institutional affil- by a sequence of three strong earthquakes [1] (Figure1) that occurred on the 3 (10:16:08 iations. UTC), 4 (18:38:19 UTC) and 12 (12:57:50 UTC) of the month. Source parameters of these earthquakes, determined by the National Observatory of Athens (NOA) as well as from other national and international seismological centers, are listed in Table S1. It is noteworthy that for the first and third strong earthquakes consistent moment magnitudes, Mw~6.3 and M ~5.7, respectively, have been determined by the various centers. However, magnitudes Copyright: © 2021 by the authors. w Licensee MDPI, Basel, Switzerland. determined for the second strong earthquake vary from Mw = 5.9 to Mw = 6.3, an issue This article is an open access article examined further later. On the other hand, the various fault-plane solutions produced distributed under the terms and by seismological centers are consistent in that the three strong earthquakes shared focal conditions of the Creative Commons mechanisms of normal faulting striking roughly NW–SE. Attribution (CC BY) license (https:// This is also consistent with the active tectonics of the area, which is dominated by a creativecommons.org/licenses/by/ NE–SW to N–S active extensional field [2–8]. The main faults that dominate the area are the 4.0/). Tyrnavos and Larissa Faults, which dip to N–NE, and the Rodia Fault, which dips to S–SW

Geosciences 2021, 11, 311. https://doi.org/10.3390/geosciences11080311 https://www.mdpi.com/journal/geosciences Geosciences 2021, 11, 311 2 of 24

Geosciences 2021, 11, 311 NE–SW to N–S active extensional field [2–8]. The main faults that dominate the area2 of are 24 the Tyrnavos and Larissa Faults, which dip to N–NE, and the Rodia Fault, which dips to S–SW (Figure 1). The Tyrnavos-Larissa Basin is a tectonic structure which is bounded by the Tyrnavos and Larissa Faults to the south and by the Rodia Fault to the north. The (Figure1). The Tyrnavos-Larissa Basin is a tectonic structure which is bounded by the TyrnavosTyrnavos and Fault Larissa is one Faults of the to major the south active and structures by the Rodia in the Fault study to area. the north. It mainly The affects Tyrnavos the FaultTriassic is one crystalline of the major limestone active of structures the Pelagonian in the studybasement area. as It mainlywell as affectsPliocene the and Triassic Qua- crystallineternary deposits limestone [2]. The of the general Pelagonian trend of basement the Tyrnavos as well Fault as Plioceneis about andE–W, Quaternary although a depositsslightly right [2]. The-bending general geometry trend of is the evident. Tyrnavos The Faultlength is of about the well E–W,-defined although and a slightlymapped right-bendingfault trace exceeds geometry 12 km is evident.[3]. Earthquake The length parameters of the well-defined determined and by mappedNOA showed fault trace that exceedsthe strong 12 kmseismic [3]. Earthquakeactivity was parameters preceded by determined a series of by about NOA 30 showed foreshocks, that thewith strong local magnitude, ML, ranging from 0.9 to 2.9, occurring in an area of radius of ~30 km around seismic activity was preceded by a series of about 30 foreshocks, with local magnitude, ML, rangingthe epicenter from 0.9of the to 2.9, first occurring shock from in an 28 area February of radius up ofto ~303 March km around (Figure the 1). epicenterThe first strong of the firstearthquake shock from caused 28 February damage up mostly to 3 March in old (Figure unreinforced1). The first houses strong and earthquake other structures caused damagemainly inmostly the villages in old unreinforced of Damasi, housesMesochori and and other Vlachogianni structures mainly [1] (Figure in the villages1). Further of Damasi,damage Mesochoriwas noted andtowards Vlachogianni NW after [the1] (Figure second1 strong). Further earthquake damage mainly was noted in the towards village NWDomeniko. after the second strong earthquake mainly in the village Domeniko.

FigureFigure 1.1. The Thessaly Thessaly area, area, Central Central Greece, Greece, that that ruptured ruptured by by three three strong strong earthquakes earthquakes on on3, 4 3,and 4 12 March 2021 with magnitudes, Mw, of 6.3, 6.2 and 5.7, respectively, as determined in this study; and 12 March 2021 with magnitudes, Mw, of 6.3, 6.2 and 5.7, respectively, as determined in this study;stars show stars strong show strongearthquake earthquake epicenters epicenters relocated relocated in the present in the present study. study.Blue and Blue purple and purplecolored circles illustrate foreshocks and aftershocks occurring from 28 February to 3 March and from 3 to 15 colored circles illustrate foreshocks and aftershocks occurring from 28 February to 3 March and March, respectively (epicentral determinations by NOA, http://www. fromgein.noa.gr/en/seismicity/earthquake 3 to 15 March, respectively (epicentral-catalogs determinations, last access 25 by May NOA, 2021 http://www.gein.noa.gr/en/). Black lines and black tri- seismicity/earthquake-catalogsangles show the main normal ,faults last access in the 25 area May [2 2021).–8] and Black settlements lines and reported black triangles in the main show text, the mainrespectively; normal faultsfault name in the codes: area [ 2L–F8 ]= and Larissa settlements Fault, T reportedF = Tyrnavos in the Fault, main R text,F = R respectively;odia Fault. In fault the nameinset map, codes: arrowsLF = Larissaillustrate Fault directio, TF =ns Tyrnavos of main lithospheric Fault, RF = motions; Rodia Fault. thick In and the thin inset rectangles map, arrows show illustratethe study directions area and ofthe main east lithosphericCorinth Gulf motions; area, respectively, thick and thin the last rectangles one discussed show the in study the main area text. and the east Corinth Gulf area, respectively, the last one discussed in the main text. The March 2021 earthquake sequence is quite challenging for a number of reasons. First,The no Marchstrong earthquake 2021 earthquake has occurred sequence in isthe quite entire challenging Thessaly area for asince number 1980, of therefore, reasons. First,the last no strongsequence earthquake is the first has recorded occurred by in themodern entire seismograph Thessaly area and since other 1980, networks, therefore, e.g. the , lastCGPS. sequence Besides, is thethe ﬁrstMarch recorded 2021 earthquakes by modern seismographruptured an area and otherwhich networks, has been e.g.,considered CGPS. Besides,as a long the-term March seismic 2021 earthquakesgap identified ruptured on the an basis area of which the historical has been seismicity considered of as the a long-termThessaly area. seismic The gap discussion identiﬁed on on thethis basisissue of opened the historical after the seismicity observation of the that Thessaly the north area. Theside discussion of Thessaly, on including this issue the opened area of after the theMarch observation 2021 activity, that thehas northnot rupture side ofd Thessaly,by strong includingearthquakes the since area of AD the 1781, March while 2021 the activity, south hasside not ruptured ruptured by by several strong strong earthquakes earthquakes since ADfrom 1781, 1954 while up to the 1980, south after side an rupturedapparent byquiescence severalstrong since AD earthquakes 1773 [9,10 from] (Figure 1954 2). up The to 1980,main afterconclusion an apparent of that quiescence apparent space since– ADtime 1773 earthquake [9,10] (Figure clustering2). The was main that conclusion should the ofseismicity that apparent pattern space–time in south Thessaly earthquake repeat clustering in the wasnorth that side, should then thea migration seismicity of pattern the ac- intivity south would Thessaly be expected repeat innorthwards the north in side, the thennear future a migration [9,10]. ofThis the important activity would issue was be expected northwards in the near future [9,10]. This important issue was further examined and debated in several subsequent studies based on geological data and on space–time seismicity patterns [11–15].

Geosciences 2021, 11, 311 3 of 24

Geosciences 2021, 11, 311 3 of 24 further examined and debated in several subsequent studies based on geological data and on space–time seismicity patterns [11–15].

FigureFigure 2. 2.Epicenters Epicenters (black (black stars), stars), dates dates and and magnitudes magnitudes of strong of strong historical historical earthquakes earthquakes that occurred that oc- incurred Thessaly in Thessaly province province and are and mentioned are mentioned in the text.in the Epicenters text. Epicenters of the of strong the strong earthquakes earthquakes of the of the 3, 4 and 12 March 2021 are shown by red (NOA’s epicenters) and blue (relocated by epicenters) 3, 4 and 12 March 2021 are shown by red (NOA’s epicenters) and blue (relocated by epicenters) stars. NOA’s epicenter of the third earthquake is not shown since it is nearly identical with the re- stars. NOA’s epicenter of the third earthquake is not shown since it is nearly identical with the located one. Positions (green triangles) and code names of the stations used for relocation are also relocatedplotted. one.Beach Positions-balls show (green the triangles)fault-plane and solutions code names of the of three the stations 2021 earthquakes used for relocation produced are by also the plotted.GEOFON/GFZ Beach-balls system. show the fault-plane solutions of the three 2021 earthquakes produced by the GEOFON/GFZ system. On the other hand, a major challenge is the investigation of the seismic fault(s) as- sociatedOn the with other the hand,three astrong major earthquakes. challenge is theThe investigationarea extended of to the the seismic south and fault(s) southeast asso- ciatedfrom the with source the three of the strong first earthquake earthquakes. is Thedominated area extended by mapped to the and south well and-studied southeast faults, fromsuch the as the source Tyrnavos of the (TF) first and earthquake Larissa is(LF) dominated ones, e.g., by [1,2,4,5] mapped (Figure and well-studied1). Α first attempt faults, to suchinvestigate as the Tyrnavos the seismotectonics (TF) and Larissa of the (LF) three ones, earthquakes e.g., [1,2,4 was,5] (Figure made 1on). Athe first basis attempt of InSAR to investigate the seismotectonics of the three earthquakes was made on the basis of InSAR analysis [16]. The authors of that study suggested that all the three earthquakes have analysis [16]. The authors of that study suggested that all the three earthquakes have been been associated with a normal fault system striking about SE–NW and dipping NE. As associated with a normal fault system striking about SE–NW and dipping NE. As we will we will see later, we propose a different seismotectonic interpretation. The only sur- see later, we propose a different seismotectonic interpretation. The only surface-fault trace face-fault trace observed in association with the March 2021 activity was found in the observed in association with the March 2021 activity was found in the area of Mesochori area of Mesochori (Figure 1) (Prof. I. Koukouvelas, personal communication) but a rele- (Figure1) (Prof. I. Koukouvelas, personal communication) but a relevant study is still in vant study is still in progress. Abundant but rather minor ground fissures were reported progress. Abundant but rather minor ground fissures were reported in the area [17]. Sur- in the area [17]. Surface manifestations of soil liquefaction were also observed in several face manifestations of soil liquefaction were also observed in several spots (Figure3 ) lying spots (Figure 3) lying within the limiting distance, R, from the epicenter of the first within the limiting distance, R, from the epicenter of the first earthquake, predicted by em- earthquake, predicted by empirical relationships between R and earthquake magnitude, pirical relationships between R and earthquake magnitude, M [18]. However, such ground M [18]. However, such ground fissures do not provide evidence of surface-fault trace fissures do not provide evidence of surface-fault trace since liquefaction is controlled by othersince factors, liquefaction such asis c M,ontrolled R and the by soilother susceptibility. factors, such as M, R and the soil susceptibility.

Geosciences 2021, 11, 311 4 of 24 Geosciences 2021, 11, 311 4 of 24

(a) (b)

FigureFigure 3. ( 3.a,b(a).,b Soil). Soil liquefaction liquefaction observed observed after after the 3 3 MarchMarch 2021 2021 M Mw w= = 6.3 6.3 earthquake earthquake near near the thePinior Piniorriver river bank, bank, a few a few km km east east of Piniada of Piniada village village (see (see location location in Figure in Figure1). 1).

TheThe present present paperpaper focuses focuses on onthe theinvestigation investigation of the of theseismotectonics seismotectonics of the of the strong strong earthquakeearthquake activity activity of of March March 2021 2021 and and the the investigation investigation of fault of fault models models for the forthree the strongest three strongestearthquakes earthquakes on the basison the of seismologicalbasis of seismological and satellite and geodesy satellite (DInSAR) geodesy observations.(DInSAR) ob- We servations.are also interestedWe are also to understandinterested to whether understand or not whether the strong or earthquakes not the strong ruptured earthquakes segments rupturedof thesame segments fault orof ofthe different same fault faults. or Ofof different equal interest faults. is theOf equal investigation interestof is possiblethe inves- links tigationbetween of possible the March links 2021 between seismogenic the March faults 2021 with seismogenic the already known faults with and mapped the already faults knownhaving and surface mapped expression faults having in the areasurface (Figure expression1). Results in the on sucharea issues(Figure are 1). of Results importance on suchfor issues the better are of assessment importance of thefor levelthe better of seismic assessment hazard of in the the level Thessaly of seismic area. hazard in the Thessaly area. 2. Materials and Methods 2. MaterialsTo investigate and Methods the seismotectonics of the three strong earthquakes that ruptured the ThessalyTo investigate area during the seismotectonics March 2021, we of relocatedthe three strong their hypocenters, earthquakes invertedthat ruptured teleseismic the ThessalyP-wave area records, during produced March Interferometric2021, we relocated Synthetic their hypocente Aperture Radarrs, inverted (InSAR) teleseismic images and P-waveinverted records, InSAR produced data, and Interferometric took also critically Synthetic into Aperture account momentRadar (InSAR) tensor images solutions and pro- invertedduced InSAR by several data seismological, and took also centers. critically Eventually into account we redetermined moment tensor moment solutions magnitudes pro- ducedand concludedby several with seismological fault models centers. for the Eventually three strong we earthquakes redetermined and moment proposed magni- a seismo- tudestectonic and concluded interpretation with for fault the models entire sequence. for the three strong earthquakes and proposed a seismotectonic2.1. Earthquake interpret Magnitudesation for the entire sequence.

2.1. ΕarthquakeConsistent Magnitudes moment magnitudes, Mw, have been determined by various seismological centers for each one of the ﬁrst and third strong earthquakes (Table S1). However, Mw determinedConsistent for moment the second magnitudes, earthquake M rangesw, have from been 5.9 determined to 6.3. It is remarkableby various seismologi- that the world- calwide centers recognized for each one Global of the Centroid first and Moment third strong Tensor earthquakes (GCMT) Project (Table andS1). However, the GEOSCOPE Mw determinedObservatory-French for the second Global earth Networkquake ranges of broad from band 5.9 seismic to 6.3. stations It is remarkable did not produce that the mo- worldwidement tensor recognized solutions Global for this Centroid earthquake. Moment In addition, Tensor the (GCMT) United Project States Geologicaland the GEO- Survey SCOPE(USGS), Observatory another renowned-French Global institute, Network has been of ablebroad to band determine seismic only stations body-wave did n mag-ot produce moment tensor solutions for this earthquake. In addition, the United States Ge- nitude mb = 5.8 for the same earthquake. On the other hand, the solution produced by ologicalthe GEOFON Survey (USGS),system of another the Deutsche renowned GeoForschungsZentrum institute, has been able(GFZ), to Potsdam, determine provided only b bodyMw-wave= 6.3 magnitudebased on regional m = 5.8 stations for the ofsame the GEOFONearthquake. system On the without other hand, the use the of solution teleseismic producedrecords. by The the reason GEOFON is that teleseismic system of records the Deutsch of thee second GeoForschungsZentrum earthquake have been (GFZ), covered Potsdam,by the recordsprovided of M anw earlier= 6.3 based large on earthquake regional (stationsMw = 7.4 of) occurringthe GEOFON in the system Kermadec–New without theZealand use of region teleseismic on 4 March records. 2021 The at 17:41:24.2 reason isUTC that (Joachim teleseismic Saul, records GFZ; personal of the communi-second

Geosciences 2021, 11, 311 5 of 24

cation). We redetermined moment magnitudes from the inversion of teleseismic records for the ﬁrst earthquake and from the inversion of InSAR data for the three strong earthquakes.

2.2. Hypocenter Relocation To improve the hypocentral determinations of NOA for the three largest earthquakes of the seismic sequence (Table S1) we relocated the hypocenters by implementing a 1D velocity model (Table1) and by employing the Non-Linear Location (NLLoc) algorithm [ 19,20], which follows a non-linear location approach [21] and provides more reliable solutions and hypocenter error estimates as compared to linearized algorithms. The velocity model used is a revision of the one found in an earlier study for a nearby area [22]. NLLoc provides a complete, probabilistic solution of the earthquake location problem expressed in terms of the posterior density function (PDF) in the space and time domains. Only manually re- picked P and S phases recorded by permanent stations of the Hellenic Unified Seismological Network (https://bbnet.gein.noa.gr/HL/real-time-plotting/husn/husnmap, last access 13 May 2021) situated at epicentral distances of up to ~120 km from the first shock were utilized. The reason is that the crustal thinning from the plate boundary towards the back- arc area creates significant errors in accurately locating the earthquake, especially when distant seismic phases are included in the analysis [23]. The difference in the onset times picked relative to NOA’s solution is on average ± 0.15 s. Following an iteration procedure, the relocation was repeated, by including phase residuals at seismic stations obtained from the previous run, until no further significant decrease of the RMS value between two successive runs was achieved, i.e., until minimizing the location errors. To correct for irregular station distribution a relevant station weighting tool in NLLoc was utilized.

Table 1. Speciﬁcations of the seismic velocity model used; Vp and Vs are velocities of P and S waves, respectively; D is layer thickness.

Vp (km/s) Vs (km/s) D (km) 2.9 1.68 0.00 4.5 2.60 1.50 5.8 3.35 4.00 6.1 3.52 7.00 6.3 3.64 12.00 6.8 3.93 18.00

2.3. Earthquake Focal Mechanisms Earthquake focal mechanisms for the three strong earthquakes have been obtained through moment tensor solutions produced by several seismological centers (Table S2). To understand similarities and differences between the various fault-plane solutions we calculated the average strike, dip and rake (slip vector) for each one of the two nodal planes (NPs) involved in the solutions published for each one of the three strong earthquakes. The results received are presented in Section 3.2.

2.4. Rupture Process from the Inversion of Teleseismic P-Waveforms The temporal and spatial evolution of the seismic slip upon the fault plane of the first shock of 3 March 2021 has been modeled applying a non-negative, least squares finite-fault inversion scheme based on a kinematic parameterization of the fault [24–26]. The data set used consists of P waveforms from teleseismic records at distances ranging from 30◦ to 90◦ with good azimuthal coverage (Figure4) and a high signal to noise ratio. Waveform data were collected from GEOFON and other seismographic networks and downloaded through the IRIS (Incorporated Research Institutions for Seismology) Data Management Center (https://ds.iris.edu/ds/nodes/dmc/, last access 20 May 2021). The waveforms were cut nearly 2 s before the first P arrival and their duration used for inversion is 25 s. All waveforms were pre-processed to remove the mean offset and instrument response before the inversion. They were also band-pass filtered between 0.04 and 0.6 Hz using Geosciences 2021, 11, 311 6 of 24

downloaded through the IRIS (Incorporated Research Institutions for Seismology) Data Management Center (https://ds.iris.edu/ds/nodes/dmc/, last access 20 May 2021). The

Geosciences 2021, 11, 311 waveforms were cut nearly 2 s before the first P arrival and their duration used for6 of in- 24 version is 25 s. All waveforms were pre-processed to remove the mean offset and instrument response before the inversion. They were also band-pass filtered between 0.04 and 0.6 Hz using a Butterworth filter, re-sampled to 0.2 samples/s and finally integrated ain Butterworth time to obtain ﬁlter, displacements. re-sampled to The 0.2 samples/s calculated and elementary ﬁnally integrated synthetics in were time convolved to obtain displacements.with an attenuation The calculated operation elementaryunder the assumption synthetics were that convolvedt* = 1 s, where with t* an is attenuationthe attenua- operationtion parameter under of the teleseismic assumption body that waves t* = that 1 s, represents where t* is the the total attenuation body wave parameter travel time of teleseismicdivided by body Q along waves the that ray representspath for P the waves. total More body wavedetails travel on the time application divided by of Q method along thefor rayearthquakes path for P in waves. the Mediterranean More details on region the application and elsewhere of method can be for found earthquakes in several in thepa- Mediterraneanpers e.g., [24,25,27] region. and elsewhere can be found in several papers e.g., [24,25,27].

FigureFigure 4. 4.Geographic Geographic distribution distribution ofof the the teleseismic teleseismic stations stations at at epicentral epicentral distances distances ( 30(30◦°< <∆ Δ< < 90 90◦°)) fromfrom where where P-wave P-wave records records of of the the 3 3 March March 2021 2021 earthquake earthquake were were used; used;beach-ball beach-ball shows shows the the earthquakeearthquake focal focal mechanism. mechanism.

ThisThis methodmethod couldcould notnot bebe appliedapplied toto thethe secondsecond strongstrong earthquakeearthquake ofof thethe4 4March March sincesince no no adequate adequate number number of of teleseismic teleseismic P-wave P-wave records records are are available. available. As As explained explained earlier, ear- thelier, reason the reason is that is that in many in many stations stations the the records records have have been been covered covered by by the the records records of of a a largelarge earthquake earthquake that that occurred occurred in in the the New New Zealand–Kermadec Zealand–Kermadec seismic seismic zone. zone. On On the the other other hand,hand, thethe earthquakeearthquake ofof 1212 MarchMarch waswas notnot largelarge enough enough toto apply apply the the method. method. Therefore,Therefore, thethe investigation investigation of of the the space space and and time time seismic seismic slip slip distribution distribution from from the the inversioninversion of of teleseismicteleseismic P-waveformsP-waveforms waswas restrictedrestricted to the first ﬁrst shock shock of of 3 3 March March 2021. 2021. TheThe seismicseismic fault dimensions were were chosen chosen large large enough enough to to permit permit the the slip slip upon upon the thefault fault in all in allpossible possible directions directions if it if is it imposed is imposed by bythe thewaveform waveform data. data. Namely, Namely, the therec- rectangulartangular fault fault created created was was of of32.5 32.5 km km in inlength length and and of 16 of 16km km in depth in depth measured measured from from the ◦ thesurface. surface. Taking Taking a fault a fault dip dipof ~45 of° ~45 the thealong along dip dipwidth width of the of fault the fault is about is about 25 km. 25 km.The Thefault fault was was discretized discretized to 180 to 180sub- sub-faultsfaults (cells), (cells), 18 of 18 them of them along along strike strike and and10 along 10 along dip. dip.Each Each point point source source response response was was computed computed with with a a code code based based on on the generalized generalized ray ray theorytheory (e.g., (e.g., [ 28[28]])) and and the the use use of of the the crustal crustal velocity velocity model model with with the the speciﬁcations specifications shown shown in Tablein Table1. We 1. adoptedWe adopted the relocatedthe relocated hypocenter hypocenter obtained obtained after after our analysis,our analysis, which which was setwas atset horizontal at horizontal distance distance of ~19 of km~19 fromkm from the southeastthe southeast edge edge of the of fault.the fault. The exactThe exact way way the syntheticsthe synthetics were were constructed constructed followed followed the discussion the discussion by [29 by]. [29]. According to the fault-plane solutions available for the 3 March 2021 strong earthquake (Table S2) the fault plane dips towards either NE or SW. However, a fault dipping NE is consistent with the active tectonics of the area (Figure1). After several inversion iterations

the strike of 312◦ and dip of 45◦ were found to better ﬁt the data for the NE-dipping fault. As we will see later, this is consistent with the source solution found after InSAR analysis too. Although fault-plane solutions indicate a rake of about −90◦, which implies nearly Geosciences 2021, 11, 311 7 of 24

pure normal fault, the rake vector was allowed to vary within the range from −60◦ to −150◦ during the inversion procedure. The procedure was performed repeatedly setting different values of rupture velocity, varying from 2.5 to 3.0 km/s. As the rupture velocity did not change during each one of the inversion iterations, six time windows were inserted with 0.3 s time lag. This option allowed a rise time of up to 1.8 s to be taken on each sub-fault, if required by the observations. Table2 summarizes the most important parameters involved in the inversion process.

Table 2. Best ﬁt model parameters for the fault plane. L: fault length, H: depth of faulting, v: rupture velocity, h: focal depth of centroid, Mo: seismic moment (N*m), Mw: moment magnitude. Dip direction of the fault is to northeast, rake is calculated at the main slip patch of the fault (see text).

Strike Dip Rake L (km) H (km) v (km/s) h (km) Mo Mw 312◦ 45◦ −95◦ 32.5 16 2.7 12 3.5 × 1018 6.3

2.5. Rupture Process from InSAR Copernicus is a European Commission initiative, which, in cooperation with the European Space Agency (ESA), developed the new family of Sentinel satellites. Sentinel-1 carries a sophisticated Synthetic Aperture Radar (SAR) instrument for capturing images of the Earth’s surface. The radar instrument uses C-band of microwave radiation and can operate in four modes, but the standard product is the Interferometric Wide Swath (IWS). Concerning the IWS polarimetry the wave has a single mode polarization (Vertical–Vertical or Horizontal–Horizontal). The default IWS mode over land has a swath width of 250 km and a ground resolution of 5 × 20 m. This mode images in three sub-swaths using the Terrain Observation with Progressive Scans SAR (TOPSAR), while the actual repeat period is 6 days. One of the main applications of the IWS mode scenes concerns the monitoring of land deformation. Sentinel-1 SAR scenes assure a series of advantages: (a) continuous, all-weather day and night imagery, (b) rapid revisit period in the same imaging mode (6 days), (c) constant and regular acquisition to build up a large global archive, (d) wide area coverage, thanks to the 250 km image swath width, and (e) narrow orbital tube. In order to detect and measure surface deformation caused by the earthquake activity during March 2021, Sentinel-1 IWS Single Look Complex (SLC) products (Table S3), cover- ing the study area were downloaded from the Copernicus Open Access Hub (https://scihub.copernicus.eu/dhus/#/home, accessed on 14 April 2021). The sens- ing dates range from 25 February to 15 March 2021 for four ascending passes and from 24 February to 14 March 2021 for three descending passes. The digital elevation model (DEM) used is based on the NASA’s Shuttle Radar Topography Mission (SRTM) 3 arc- seconds DEM (https://srtm.csi.cgiar.org/10, accessed on 14 April 2021), which is of spatial resolution of approximately 90 m/pixel. Spaceborne SAR interferometry produces 3D topographic data of the Earth’s surface directly from two SAR images [30]. An extension of the basic technique, called Differential SAR Interferometry (DInSAR), allows measurements of land deformation [31–33]. DInSAR exploits the phase difference between two or more coherent complex-valued images, i.e., SLC products, in order to derive path-length differences in the scale of the carrier wavelength and below. The capability of SAR interferometry to remotely monitor areas much wider than traditional surveying techniques, makes this technique particularly suitable for both regional and local scales [34]. In the case of Sentinel-1, DInSAR processing presents some peculiarities because of the special TOPSAR mode used by the SAR sensor to acquire the Interferometric Wide-Swath (IWS) data. The S1 IWS SLC product, used for interferometric applications, consists of three sub-swaths and each sub-swath image consists of a series of bursts. What we needed for our analysis was (i) the minimum of two SAR images taken before (master scene) and after (slave scene) the earthquake event and forming an interferometric pair, and (ii) a DEM Geosciences 2021, 11, 311 8 of 24

of the area under study. Table S4 lists features of the master and slave scenes used in the present study. The main processing steps followed are the following: co-registration of the two complex images, topography elimination using the DEM of the area, baseline estimation, generation of the interferogram, DEM/ellipsoid interferogram flattening, adaptive fil- tering/estimation and generation of the interferometric coherence, phase unwrapping, phase to displacement. These steps were performed with the use of the ENVI SARscape® software (L3Harris Geospatial, Boulder, CO, USA). After applying the above main steps, DInSAR was only able to measure the path length difference in its Line of Sight (LoS) direction. Therefore, DInSAR-derived displacements represent the 1D deformation along the LoS direction. LoS represents the direction/distance between the SAR sensor and the target at hand. This is an essential and general limitation of SAR systems. By using ascending and descending LoS deformation maps we decomposed LoS displacements [35,36] into vertical (up–down) and horizontal (east–west) deformation maps. After obtaining the shapefile of sampled points from the raster DInSAR map, the coseismic signal was modeled through Non-Linear and Linear inversions [37], thus allowing to infer the geometry, kinematics and slip distribution in the seismic fault. The displacement model and the determination of a single fault with distributed slip are produced by following several steps. First an image sampling is required, using the LoS displacement, i.e., the observed displacement, to set a number of points to model. The image sampling was carried out using the equally spaced points approach. To produce the surface displacement, which is induced by a rectangular fault plane in a homogeneous and elastic half-space, a well-known elastic dislocation approach [38] was followed through Non-Linear Inversion. The best-fit source is determined by a set of fault parameters (dip, rake, strike, length), which better predict the observed displacement data. After finding the best-fit solution, inversion is carried out to determine the slip distribution over the fault plane constrained via Non-Linear inversion. The above procedure was applied sequentially for the three strong earthquakes under investigation.

3. Results 3.1. Hypocentral Relocation The relocated hypocenters of the three strongest earthquakes of the seismic sequence are listed in Table3 and displayed in Figure2. As regards the ﬁrst and second earthquakes the new epicenters are shifted towards SW about 2.5 and 1.5 km away from the respective NOA’s epicenters. For the third earthquake, however, the relocated and the NOA’s epicenters are very close, which is due to the fact that NOA’s routine determinations are improved thanks to the installation and operation of six portable seismograph stations by the Geophysical Laboratory, Aristotelian University of Thessaloniki, before the occurrence of the third earthquake.

Table 3. Relocated hypocenters for the three strong earthquakes.

o o ◦ Date Time (UTC) Lat (ϕ N) Long (λ E) Depth (km) Mw Gap Erz (km) Erh (km) DD.MM.YYYY HH:MM:SS 3.3.2021 10:16:08 39.7417 22.1854 10.5 6.3 59 1.0 0.68 4.3.2021 18:38:19 39.7869 22.1168 4.0 6.2 56 0.67 0.76 12.3.2021 12:57:50 39.8374 22.0114 4.5 5.7 41 0.5 0.54

The RMS found for the three relocated hypocenters are less than 0.22, while in NOA’s solutions it is equal to 0.67, 0.47 and 0.41 for the 1st, 2nd and 3rd strong earthquakes, respectively. It is noteworthy that according to the calculated focal depths, h, the second and third earthquakes were found quite shallow, with h = 4.0 km and h = 4.5 km, respectively, as compared to the 3 March ﬁrst shock (h = 10.5 km). The respective focal depths determined Geosciences 2021, 11, 311 9 of 24

by NOA are 8.5 km, for the ﬁrst shock, and 4.8 and 7.0 km for the next two earthquakes. However, the centroid depths found from the inversion of teleseisic records (Table2) and of InSAR data (Section 3.3) as well as from the moment tensor solutions produced by the various centers (Table S1) are in general larger than the focal depths particularly of the second and third earthquakes.

3.2. Earthquake Focal Mechanisms From the results summarized in Table4, it is revealed that the solutions produced for each one of the three earthquakes are consistent each other since the average strikes, dip and rake have small standard deviations. Comparing the three earthquakes, we found that they are characterized by a strike which is about NW–SE or WNW–ESE. On the other hand, both nodal planes (NPs) of the ﬁrst two earthquakes have nearly the same average dip, which ranges between 44◦ and 49◦ but it is either 39◦ or 54◦ for the third earthquake. The values of the average rake imply that the faulting style of the three earthquakes is nearly pure normal with a small strike-slip component. However, the strike-slip component is relatively larger in the third earthquake.

Table 4. Average (Av) geometric features of the nodal planes (NPs) determined in fault-plane solutions produced by several national and international seismological centers (see Table S2) for the three strong earthquakes.

◦ ◦ ◦ Date Time NP Av Strike ( )Av Dip ( )Av Rake ( ) DD.MM.YYYY HH:MM:SS 1 130 ± 9 49 ± 7 −93 ± 11 3.3.2021 10:16:08 2 314 ± 7 48 ± 8 −88 ± 12 1 125 ± 13 46 ± 10 −91 ± 2 4.3.2021 18:38:19 2 303 ± 16 44 ± 10 −90 ± 2 1 106 ± 9 39 ± 7 −105 ± 16 12.3.2021 12:57:50 2 304 ± 11 54 ± 8 −80 ± 11

3.3. Seismic Slip Distribution The synthetics obtained for the 3 March 2021 ﬁrst shock ﬁt well-enough the recorded waveforms (Figure5). The heterogeneous spatial slip distribution, which is illustrated in a SE–NW cross-section of the fault area (Figure6), shows that seismic slip occurred in a main patch of the fault plane and in a secondary patch situated to the southeast but at shallower depth with respect to the main one. The main slip area is concentrated at depths from 6 to 15 km with the maximum slip of ~1.20 m concentrated at depths of 10–12 km. The minimum slip of ~20 cm occurred near the surface. The rupture length, L, changes with depth, being L~15 km at depths from 8 to 15 km and L~22 km from below the surface down to 8 km. The increase of L to the upper part of the fault plane is due to the fact that the secondary slip patch was developed at shallower depths with respect to the main patch. From the inversion procedure we found that the rake values upon the fault plane varies between −80◦ and −115◦ but the mean rake value in the main patch of slip is about −95◦, which is consistent with the average rake of −93◦ (Table4) calculated from the various fault-plane solutions (Table S2). Geosciences 2021, 11, 311 10 of 24 Geosciences 2021, 11, 311 10 of 24

Figure 5.5. ScaledScaled fitfit between between real real waveforms waveforms (blue (blue lines) lines and) and synthetics synthetics (red (red lines) lines) forthe for30 the stations; 30 stations; station station name name (e.g., TIXI(e.g.),, TIXI), station epicentral distance in degrees (Dis.) and azimuth (Az.) are also shown. station epicentral distance in degrees (Dis.) and azimuth (Az.) are also shown. × 18 The total seismic moment moment released released was was estimated estimated at at M Moo == 3.53.5 × 101018 N*m, which corresponds to M w == 6 6.3..3. The The moment moment rate rate function function shows shows that that the total duration of the seismic rupturerupture processprocess was was ~10.2 ~10.2 s buts but the the main main moment moment release release occurred occurred within within the firstthe first4 s of 4 ruptures of rupture (Figure (Figure7). From 7). From the timethe time evolution evolution of seismic of seismic slip, slip, which which is illustrated is illustrated by bysix six sequential sequential snapshots snapshots with with time time step step of of nearly nearly 1.7 1.7 s s (Figure (Figure6 ),6), it it is is evident evident that that afterafter the initiation of of the process the rupture propagated mainly upwards for about 7 s. After reaching close to the surface the rupture propagated bilaterally but mainly towardstowards SE where the secondary patch of slip developed. where the secondary patch of slip developed. The horizontal projection of the rupture area of the first shock (Figure8) shows that The horizontal projection of the rupture area of the first shock (Figure 8) shows that the most affected villages from the first earthquake, like Mesochori and Damasi [1], are the most affected villages from the first earthquake, like Mesochori and Damasi [1], are situated within the rupture area. On the contrary, in the town of Tyrnavos and in the city situated within the rupture area. On the contrary, in the town of Tyrnavos and in the city of Larissa, which are situated outside the rupture area, less damage was reported. It is also of Larissa, which are situated outside the rupture area, less damage was reported. It is evident that most of the foreshocks that preceded the first shock from the 28 February to also evident that most of the foreshocks that preceded the first shock from the 28 Febru- the 3 March 2021 occurred within the main asperity that ruptured. ary to the 3 March 2021 occurred within the main asperity that ruptured.

Geosciences 2021, 11, 311 11 of 24 Geosciences 2021, 11, 311 11 of 24 Geosciences 2021, 11, 311 11 of 24

Figure 6. Snapshots showing the space–time evolution of slip upon the fault which◦ is striking at Figure 6. Snapshots showingFigure the space–time 6. Snapshots evolution showing of slip the upon space the–time fault evolution which is striking of slip atupon 138 theand fault dipping which northeast; is striking at 138° and dipping northeast; star denotes the hypocenter, fault plane dimension is in km, slip con- star denotes the hypocenter,138° fault and plane dipping dimension northeast; is in km, star slip denotes contours the hypocente are in m. Slipr, fault amplitude plane dimension in m is shown is inin km, color slip contours are in m. Slip amplitude in m is shown in color and the motion direction, i.e., the rake angle, and the motion direction, i.e.,tours the are rake in angle,m. Slip of amplitude the hanging in m wall is shown relative in to color the footwall,and the motion is indicated direction, with i.e. red, the arrows rake angle, of the hanging wall relative to the footwall, is indicated with red arrows showing direction and showing direction and amplitudeof the of hanging slip. wall relative to the footwall, is indicated with red arrows showing direction and amplitude of slip. amplitude of slip.

Figure 7. Moment rate function of the 3 March 2021 earthquake of Mw = 6.3. FigureFigure 7. 7.Moment Moment raterate functionfunction ofof thethe 3 3 March March 2021 2021 earthquake earthquake of of M Mww == 6.3.6.3.

Geosciences 2021, 11, 311 12 of 24 Geosciences 2021, 11, 311 12 of 24

FigureFigure 8. 8.Horizontal Horizontal projectionprojection ofof the the slip slip distribution distribution for for the the 3 3 March March rupture. rupture. Stars Stars from from SE SE to to NWNW show show relocated relocated epicenters epicenters of of the the ﬁrst, first, second second and and third third earthquakes; earthquakes; magnitudes magnitudes determined determined in this study (see later sections). Black circles illustrate foreshocks recorded from 28th February to in this study (see later sections). Black circles illustrate foreshocks recorded from 28th February to 3 March 2021; epicenters are from NOA’s determinations (http://bbnet.gein.noa.gr/HL/databases/ 3 March 2021; epicenters are from NOA’s determinations (http://bbnet.gein.noa.gr/HL/databases/ database, accessed on 15 May 2021). database, accessed on 15 May 2021).

3.4.3.4. Ground Ground Deformation Deformation from from INSAR INSAR TheThe results results obtained obtained from from InSAR InSAR analysis analysis are characterized are characterized by coherent by coherent values ranging values fromranging 0.0 tofrom 1.0, 0.0 the to value 1.0, the 1 corresponding value 1 corresponding to very high to very coherence. high coherence. The DInSAR The processDInSAR producedprocess produced images in images radian unitin radian in the unit range in −theπ torangeπ, which −π to caused π, which ambiguity caused problems.ambiguity Althoughproblems the. Although pattern ofthe deformation pattern of deformation could be detected, could be the detected, main information the main information regarding theregarding deformation the deformation value could value not be could read properly.not be read In orderproperly. to get In results order ofto deformationget results of containingdeformation metric containing value, an metric unwrapping value, an process unwrapping was performed, process was and theperformed phase unit, and was the transformedphase unit was into transformed metric units ininto LoS metric for every units interferometric in LoS for every pair. interferometric This unwrapping pair. step This wasunwrapping undertaken step with was the undertaken use of the with Minimum the use Cost of the Flow Minimum method [Cost39]. Flow method [39].

3.4.1.3.4.1. The The First First Shock Shock of of 3 3 March March 2021 2021 TheThe wrapped wrapped interferogram interferogram ofof thethe 3 3 March March earthquake earthquake andand the the corresponding correspondingdis- dis- placementplacement map map (Figure (Figure9 )9) based based on on the the interferometric interferometric pair pair are are composed composed by by Sentinel Sentinel 1 1 SLCSLC images images taken taken in ascendingin ascending mode mode on 25 on February 25 February 2021 and2021 3 Marchand 3 March 2021 for 2021 the master for the andmaster slave and images, slave respectively. images, respectively. A clear pattern A clear of 14 pattern fringes of is 14evident, fringes which is evident, forms a which lobe showingforms a lobe subsidence. showing From subsidence. the displacement From the displacement map the subsidence map the amplitude subsidence is estimatedamplitude upis estimated to about − up40 to cm. about The −4 subsidence0 cm. The subsidence domain is nearlydomain identical is nearly with identical the main with patchthe main of rupturepatch of obtained rupture fromobtained the inversionfrom the inversion of P waveforms of P waveforms (Figure8). (Fig Inure the 8). southern In the southern part of thepart lobe of the of fringeslobe of fringes an incomplete an incomplete fringe isfringe also is recognized also recognized corresponding corresponding to uplift, to uplift, the amplitudethe amplitude of which of which from thefrom displacement the displacement map is map estimated is estimated up to ~3 up cm. to The~3 cm. comparison The com- ofparison LoS displacements of LoS displacements obtained fromobtained the observedfrom the andobserved modeled and datamodeled are nearly data are identical nearly givenidentical that given onlyminimal that only residuals minimalwere residuals received were (Figure received 10 ).(Figure 10).

Geosciences 2021, 11, 311 13 of 24

Geosciences 2021, 11, 311 13 of 24 Geosciences 2021, 11, 311 13 of 24

((aa) ) (b) (b)

FigureFigureFigure 9. 9. 9. ((aa))( a Wrapped Wrapped) Wrapped interferogram interferogram interferogram and and and (b ) ( (bb LoS)) LoS LoS displacement displacement displacement for for the for the first the ﬁrst seismic first seismic seismic event event of event 3 of of 3 MarchMarch3 March 2021 2021 2021 fromfromfrom master master and and and slave slave slave images images images taken taken in ascending inin ascendingascending mode mode mode on on25 on 25February 25 February February 2021 2021 and 2021 and 3 and 3 March 2021, respectively; star shows relocated epicenter. March3 March 2021, 2021, respectively; respectively; star star shows shows relocated relocated epicenter.epicenter.

Figure 10. Comparison of LoS displacements from observed and modeled data for the first earthquake (3 March 2021) using the ascending mode. FigureFigure 10. 10. ComparisonComparison ofof LoSLoS displacements displacements from from observed observed and modeled and modeled data for thedata ﬁrst for earthquake the first earthquake(3 March The(3 March best 2021)-fit 2021) using solution using the ascending found the ascending for mode. the source mode. (Table 5, Figure 11) is in general consistent with the solution determined from the inversion of teleseismic P waveforms. The geo- The best-ﬁt solution found for the source (Table5, Figure 11) is in general consistent deticThe source best solution-fit solution provides found a seismicfor the faultsource dipping (Table to 5, NE Figure with 11)strike is inSE general–NW (317 consistent°), with the solution determined from the inversion of teleseismic P waveforms. The geodetic withdip angthe lsolutione of 30° anddetermined rake −110 from°. The the geodetic inversion seismic of teleseismic moment (Table P waveforms. 5) calculated The geo- source solution provides a seismic fault dipping to NE with strike SE–NW (317◦), dip through the Okada formalism analyzed in [38] results in earthquake magnitude Mw = 6.3, deangletic source of 30 ◦solutionand rake provides−110◦. The a seismic geodetic fault seismic dipping moment to (TableNE with5) calculated strike SE through–NW (317°), which coincides with the seismically determined magnitude. The maximum slip found is dipthe ang Okadale of formalism 30° and analyzed rake −110 in° [.38 The] results geodetic in earthquake seismic magnitude moment M (Table= 6.3, 5) which calculated 1.5 m at depth 5–7 km, which is shallower than the depth of 10–12 km determinedw from throughcoincides the with Okada the seismicallyformalism determined analyzed in magnitude. [38] results The in maximum earthquake slip foundmagnitude is 1.5 m M atw = 6.3, the seismic inversion method. This is probably due to that the geodetic fault dip angle is whichdepth coincides 5–7 km, which with isthe shallower seismically than determined the depth of 10–12magnitude. km determined The maximum from the slip seismic found is smaller than the one (45°) found from both the seismic inversion method and the average 1.5inversion m at depth method. 5–7 km, This which is probably is shallower due to that than the geodeticthe depth fault of dip10– angle12 km is determined smaller than from angle (48°) of◦ the various fault plane solutions published (Table 4). ◦ thethe seismic one (45 inversion) found from method. both the This seismic is probably inversion due method to that and the the geodetic average angle fault (48dip) angle of is the various fault plane solutions published (Table4). smallerTable 5. than Seismic the source one (45solutions°) found determined from both from InSARthe seismic analysis inversion for the three method strong earthquakes and the average Table 5. Seismic sourceangleof solutions March (48 °2021.) determined of the various from InSAR fault analysisplane solutions for the three published strong earthquakes (Table of4). March 2021.

◦ ◦ ◦ 18 Moment Maximum Date Time StrikeDate ( ) DipTime ( ) S Raketrike ((°)) DMomentip (°) R (×ake10 (°N*m)) M MaximumMw Slip (m) Table 5. Seismic source solutions determined from InSAR analysis(×1018 N*m) forw the three stSronglip (m) earthquakes DD.MM.YYYY HH:MM:SSofDD.MM.YYYY March 2021. HH:MM:SS 3.3.2021 10:16:083.3.2021 317 10:16:08 30 −317110 30 3.02−110 3.02M 6.3 oment6.3 1.5 1.5 Maximum 4.3.2021 18:38:194.3.2021D 116ate 18:38:19 40Time S−trike116101 (°) 40D ip (°) 2.08−1R01ake (°) 2.08 6.2 6.2 M 1.2w 1.2 (×1018 N*m) Slip (m) 12.3.2021 12:57:5012.3.2021 117 12:57:50 45 −11795 45 0.37−95 0.37 5.7 5.7 1.0 1.0 DD.MM.YYYY HH:MM:SS 3.3.2021 10:16:08 317 30 −110 3.02 6.3 1.5 4.3.2021 18:38:19 116 40 −101 2.08 6.2 1.2 12.3.2021 12:57:50 117 45 −95 0.37 5.7 1.0

Geosciences 2021, 11, 311 14 of 24 Geosciences 2021, 11 , 311 14 of 24 Geosciences 2021, 11, 311 14 of 24

(a) (b) (a) (b) Figure 11. Modelled geodetic seismic fault for the first shock of 3 March 2021 in (a) 3D illustration Figure 11. ModelledFigure 11. geodetic Modelledand seismic ( bgeodetic) vertical fault seismic profile for the faultalong ﬁrst shockfor strike the of first(SE 3 March– NW).shock 2021 of 3 inMarch (a) 3D 2021 illustration in (a) 3D and illustration (b) vertical proﬁle along strikeand (SE–NW). (b) vertical profile along strike (SE–NW). 3.4.2. The Earthquake of 4 March 2021 3.4.2. The Earthquake of 4 March 2021 3.4.2.For The the Earthquake second strong of 4 Marchseismic 2021 event of 4 March, the wrapped interferogram and the For the second strong seismic event of 4 March, the wrapped interferogram and the displacementFor the secondmap (Figure strong 12) seismic are based event on of the 4 March, interferometric the wrapped pair composed interferogram by Sentinel and the displacement map (Figure 12) are based on the interferometric pair composed by Sentinel 1displacement SLC images taken map (Figure in ascending 12) are mode based on on 3 the March interferometric 2021 and 9 pair March composed 2021 for by the Sentinel master 1 1 SLC images andtakenSLC slave images in ascending images, taken respectively.inmode ascending on 3 March Themode deform2021 on 3and Marchation 9 March pattern 2021 2021 and is 9for illustrated March the master 2021 by for four the fringes, master and slave images,whichand slaverespectively. form images, a lobe Theshowing respectively. deform groundation The subsidence pattern deformation is illustratedof estimated pattern by is amplitude illustratedfour fringes, up by to four about fringes, −10 which form a cm.lobewhich The showing form comparison a lobeground showing of subsidenceLoS grounddisplacements of subsidence estimated obtained of amplitude estimated from the up amplitude observed to about up and −1 to0 modeled about −10 data cm . cm. The comparisonareThe identical comparison of LoS given displacements of that LoS zero displacements residualsobtained fromwer obtainede thereceived observed from (Figure the and observed 13).modeled The and earthquake data modeled dataepicen- are are identical giventidenticaler falls that at givenzero the northresiduals that zero-eastern wer residuals emargin received were of (Figure receivedthe subsidence 13). (Figure The domain,13earthquake). The earthquakewhich epicen- is consistent epicenter with falls ter falls at thethe atnorth thesmall- north-easterneastern focal margindepth margin of of the the earthquake. of subsidence the subsidence However, domain, domain, which the average which is consistent is centroid consistent with depth with of the ~13 small km the small focalisfocal depth projected depth of the well of earthquake. the inside earthquake. the However,subsid However,ence the domain theaverage average. centroid centroid depth depth of of~13 ~13 km km is projected is projected wellwell inside inside the the subsid subsidenceence domain domain..

(a) (b) (a) (b) FigureFigure 12. 12. (a()a Wrapped) Wrapped interferogram interferogram and and (b) (LoSb) LoS displacement displacement for the for second the second seismic seismic event event of 4 Figure 12. (a) Wrapped interferogram and (b) LoS displacement for the second seismic event of 4 Marchof 4 March 2021 from 2021 frommaster master and slave and images slave images taken in taken ascending in ascending mode on mode 3 March on 32021 March and 20219 March and March 2021 from2021, master respectively; and slave star images shows taken relocated in ascending epicenter mode. on 3 March 2021 and 9 March 2021, respectively;9 March star shows 2021, relocated respectively; epicenter star shows. relocated epicenter.

Geosciences 2021, 11, 311 15 of 24

GeosciencesGeosciences 20212021, 11, ,11 311, 311 15 15of of24 24

Figure 13. Comparison of LoS displacements from observed and modeled data for the second earthquake (4 FigureMarchFigure 2021) 13. 13. Comparison usingComparison the ascending of ofLoS LoS displacements mode. displacements from from observed observed and and modeled modeled data data for for the the second second earthquakeearthquake (4 (4March March 2021) 2021) using using the the ascending ascending mode. mode. The best-fit geodetic source solution (Table 5, Figure 14) provides a seismic fault dipping to about TheSWThe with best best-ﬁt- fitstrike geodetic geodetic ESE– WNW source source (116 solution solution°), dip (Table angle (Table of 5,5 40, Figure Figure° and 14) rake14 ) provides provides−101°. The a a seismic seismic fault fault geodetic faultdipping solutiondipping to is toabout consistent about SW SW with with strike one strike ofESE ESE–WNWthe–WNW two average (116 (116°),◦ NPs),dip dip anglefound angle of from of 40 40° ◦variousandand rake rake −1−01101°. ◦The. The seismic faultgeodetic-planegeodetic solutions fault fault solution solution(Table is 4). consistent is The consistent geodetic with with seismicone one of ofthe moment the two two average (Table average NPs5) NPsresults found found in from from various various earthquake magnitudeseismicseismic fault fault-planeM-wplane = 6.2. solutionsSeismic solutions slip (Table (Tableon the 4).4 fault).The The geodeticoccurred geodetic seismicat seismic depths moment momentranging (Table from (Table 5)5 )results results in in 5 to 20 km butearthquake earthquakethe maximum magnitude magnitude slip of M1.2 Mw =mw 6 =was.2. 6.2. Seismic found Seismic slipat slipdepth on onthe ~7 the fault km. fault occurred occurred at atdepths depths ranging ranging from from 5 to5 to20 20km km but but the the maximum maximum slip slip of of1.2 1.2 m mwas was found found at atdepth depth ~7 ~7 km. km.

(a) (b) (a) (b) Figure 14. ModelledFigure 14. geodetic Modelled seismic geodetic fault seismic for the fault earthquake for the earthquake of 4 March 2021of 4 March in (a) 3D 2021 illustration in (a) 3D andillustration (b) vertical profile along strikeand (ESE–WNW). (b) verticalFigure profile 14. along Modelled strike geodetic (ESE–WNW). seismic fault for the earthquake of 4 March 2021 in (a) 3D illustration and (b) vertical profile along strike (ESE–WNW). 3.4.3. The Earthquakes of 3 and 4 March 2021 Combined 3.4.3. The Earthquakes of 3 and 4 March 2021 Combined Displacement3.4.3. The decomposition Earthquakes includingof 3 and 4 bothMarch the 2021 seismic Combined events of 3 and 4 March Displacement decomposition including both the seismic events of 3 and 4 March 2021 2021 was also produced from interferometric pairs in ascending mode, taken on 25 Feb- wasDisplacement also produced decomposition from interferometric including pairs both in the ascending seismic mode, events taken of 3 andon 25 4 February March ruary 2021 and 9 March 2021 for master and slave images, and in descending mode taken 20212021 was and also 9 Marchproduced 2021 from for masterinterferometric and slave pairs images, in ascending and in descending mode, taken mode on 25 taken Feb- on on 24 Februaryruary24 2021 February 2021 and and 8 March 20219 March and 2021 2021 8 Marchfor for master master 2021 and for and masterslave slave images. andimages, slave The and images. wrapped in descending The inter- wrapped mode interfer-taken ferograms clearlyonograms 24 showFebruary clearly the 2021migration show and the 8 ofMarch migration the ground 2021 of forthe deformation master ground and deformation towardsslave images. NW towards (FigureThe wrapped NW (Figure inter- 15). 15). The oppositeferogramsThe subsidence/uplift opposite clearly subsidence/uplift show pattern the migration in the pattern two of theearthquakes in theground two earthquakesdeformation is evident in is towards the evident ver- NW in the (Figure vertical tical displacement15).displacement The map opposite (Figure map subsidence/uplift 16a), (Figure which 16 verifiesa), which pattern that verifies the in thefirst that two normal the earthquakes first fault normal dips is toward faultevident dips in toward the ver- NE NE while thetical secondwhile displacement the dips second towards map dips (FigureSW. towards Τhe 16a), SW.two which The earthquakes two verifies earthquakes arethat also the are firstcharacterized also normal characterized fault by dips by toward opposite opposite polarityNEpolarity wh in ilethe the in horizontal the second horizontal dipsdisplacement towards displacement (FigureSW. Τhe (Figure 16b). two 16 earthquakesb). are also characterized by opposite polarity in the horizontal displacement (Figure 16b).

Geosciences 2021, 11, 311 16 of 24 Geosciences 2021, 11, 311 16 of 24 Geosciences 2021, 11, 311 16 of 24

(a) (b) Figure 15. Wrapped interferograms for both the 3 and 4 March 2021 earthquakes: (a) ascending and FigureFigure 15. 15. WrappedWrapped interferograms interferograms for for both both the the3 and 3 and 4 March 4 March 2021 2021 earthquakes: earthquakes: (a) ascending (a) ascending and (b) descending modes. Stars show the relocated epicenters; magnitudes according to our geodetic (band) de (scendingb) descending modes. modes. Stars show Stars the show relocated the relocated epicenters; epicenters; magnitudes magnitudes according according to our geodetic to our determinations. geodetic determinations.

(a) (b) FigureFigure 16. 16. DisplacementsDisplacements in (a) verticalvertical andand (b(b)) east–west east–west directions directions for for both both the the 3 and 3 and 4 March 4 March 2021 2021earthquakes. earthquakes. Stars Stars and and magnitudes magnitudes as in as Figure in Figure 15. 15.

TheThe solutions solutions obtained imply imply that that with with the the second second earthquake earthquake the the activated activated fault fault was wassubparallel subparallel and and antithetic antithetic to the to one the activated one activated with the with ﬁrst the earthquake. first earthquake. This is a This reminder is a reminderof similar of seismotectonics similar seismotectonics characterizing characterizing the strong the earthquake strong earthquake sequence sequence that ruptured that rupturedthe eastern the Gulfeastern of CorinthGulf of Corinth on 24 and on 2524 Februaryand 25 February and again and on again 4 March on 4 1981March in 1981 Central in CentralGreece Greece [40,41] to[40,41] the south to the of south Thessaly of Thessaly (Figure1 (Figure). This case1). This is discussed case is discussed further later. further later. 3.4.4. The Earthquake of 12 March 2021 3.4.4. The E groundarthquake deformation of 12 March caused 2021 by the third seismic event of 12 March 2021 is illustrated in wrapped interferograms, in LoS displacement maps as well as in vertical The ground deformation caused by the third seismic event of 12 March 2021 is il- and east–west decomposed displacement maps (Figures 17–19). A clear pattern of three lustrated in wrapped interferograms, in LoS displacement maps as well as in vertical and fringes forming a relative small lobe showing subsidence is evident (Figure 17). From east–west decomposed displacement maps (Figures 17–19). A clear pattern of three the displacement maps the ground subsidence is estimated up to about −10 cm, while fringes forming a relative small lobe showing subsidence is evident (Figure 17). From the from the decomposition displacement the subsidence amplitude is estimated up to −9 cm. displacement maps the ground subsidence is estimated up to about −10 cm, while from The comparison of LoS displacements found from the observed and modeled data shows the decomposition displacement the subsidence amplitude is estimated up to −9 cm. The compasmallrison residuals of LoS at the displacements southern part found of the from deformation the observed domain and obtained modeled from data the showsascend- compaing moderison (Figure of LoS 20 displacements). In the southern found part from of the lobeobserved of fringes and modeled in the ascending data shows pair

Geosciences 20212021,, 1111,, 311311 1717 ofof 2424

smallsmall residuals residuals at at the the southern southern part part of of the the deformation deformation domain domain obtained obtained from from the the asas- cending mode (Figure 20). In the southern part of the lobe of fringes in the ascending pair (Figure(Figure 17) it is also recognized an incomplete fringe corresponding to ground uplift, which from displacement maps is measured up to 2 cm. From the decomposition dis- Geosciences 2021, 11, 311 17 of 24 placement (Figure 18) it is measured up to 3 cm of uplift,, but no displacement is observed inin the the west–east direction (Figure 19). However, the residuals are equal to zero at the descendinging modemode (Figure(Figure 21).21). ForFor thisthis reasonreason,, thisthis LoSLoS displacementdisplacement isis aa preferredpreferred oneone and(Figure indicates 17) it subsidence is also recognized of ~5 cm an in incomplete the central fringe side of corresponding the deformat toiion ground field and uplift, upli whichftft ofof ~2.5from cm displacement in the NE side maps of the is measureddeformed uparea. to 2 cm. From the decomposition displacement (FigureAs observed18) it is measured also in the up case to 3of cm the of 4uplift, March but 2021 no earthquake, displacement the is epicent observeder of in the the thirdthirdwest–east earthquakeearthquake direction fallsfalls (Figure atat thethe north19north). However,--eastern margin the residuals of the aresubsidence equal to domain. zero at the This descending is again consistentmode (Figure with 21 the). Forsmall this reloca reason,tedted this((h = LoS4.5.5 km)km) displacement focalfocal depthdepth is ofof a preferred thethe earthquake.earthquake. one and However,However, indicates thethesubsidence averageaverage centroidcentroid of ~5 cm depthdepth in the of centralof ~9~9 kmkm side fallsfalls of wellwell the deformation insideinside thethe subsidencesubsidence ﬁeld and uplift domain.domain. of ~2.5 cm in the NE side of the deformed area.

((a)) ((b)) Figure 17. Earthquake of 12 March 2021: Wrapped interferograms in (a) ascending (9 March 2021 FigureFigure 17. 17. EarthquakeEarthquake of of 12 12 March March 2021: 2021: Wrapped Wrapped interferograms interferograms in in (a (a) )ascending ascending (9 (9 March March 2021 2021 master and 15 March 2021 slave image) and (b) descending (8 March 2021 master and 14 March mastermaster and and 15 15 March March 2021 slaveslave image)image)and and ( b(b) descending) descending (8 ( March8 March 2021 2021 master master and and 14 March14 March 2021 2021 slave image) mode; star illustrates the relocated epicenter. slave image) mode; star illustrates the relocated epicenter.

((a)) ((b)) FigureFigure 18. 18. EarthquakeEarthquake of of 12 12 March March 2021: 2021: (a ()a) )LoSLoS LoS dd displacementisplacementisplacement inin in ascendingascending ascending modemode mode ((9 (9 March March 2021 2021 mastermaster and and 15 15 March March 2021 2021 slave slave image), image), (b (b)) )LoSLoS LoS didi displacementsplacementsplacement inin in descendingdescending descending modemode mode ((8 (8 March March 2021 2021 master and 14 March 2021 slave image); star illustrates the relocated epicenter. mastermaster and and 14 14 March March 2021 2021 slave slave image); image); star star illustrates illustrates the the relocated relocated epicenter. epicenter.

GeosciencesGeosciencesGeosciences 20212021 2021,, 1111, ,11, 311311, 311 1818 18 ofof of2424 24 Geosciences 2021, 11, 311 18 of 24

((aa)() a ) ((bb)() b )

FigureFigureFigureFigure 19.19. 19. 19. EarthquakeEarthquake EarthquakeEarthquake ofof of 1212 of 12 March 12March March March 2021:2021: 2021: 2021: ((aa )() a (verticalverticala) )vertical vertical displacement,displacement, displacement, displacement, ((bb )() (b ebe)astast) e east–westast––ww–estwestest displacement;displacement; displacement; displacement; starstar star star illustratesillustratesillustratesillustrates relocatedrelocated relocated relocated epicenterepicenter epicenter epicenter... .

FigureFigureFigureFigure 20. 20. 20. ComparisonComparison Comparison of of of of LoS LoS LoS LoS displacements displacements displacements from from from observed observed observed observed and and and and modeled modeled modeled modeled data data data data for for for the for the the third the third third thirdearth- earthquakeearthquakeearthquakequake (12 (12(12 March (12 MarchMarch March 2021) 2021)2021) 2021) using usingusing using the thethe ascending the ascendingascending ascending mode. mode.mode. mode.

FigureFigureFigureFigure 21. 21. 21. ComparisonComparison Comparison of of of of LoS LoS LoS LoS displacements displacements displacements from from from observed observed observed observed and and and and modeled modeled modeled modeled data data data data for for for the for the thethird the third third thirdearth- earthquakeearthquakeearthquakequake (12 (12(12 March (12 MarchMarch March 2021) 2021)2021) 2021) using usingusing using the thethe descending the descendingdescending descending mode. mode.mode. mode.

TheTheTheAs bestbest observedbest--fitfit-fit geogeo geodd alsoeedtticeict icinsourcesource source the case solutionsolution solution of the forfor 4for the Marchthe the earthquakeearthquake earthquake 2021 earthquake, ofof of 1212 12 MarchMarch March the 20212021 epicenter 2021 (Table(Table (Table of 5,5, the 5, FigureFigureFigurethird 22)22) earthquake 22) providesprovides provides falls aa seismicaseismic atseismic the north-easternfaultfault fault dippingdipping dipping margintoto to aboutabout about of SWSW the SW with subsidencewith with strikestrike strike domain.ESEESE ESE––WW–WNWNW ThisNW (117(117 is(117 again°°),),° ), dipdipdipconsistent angleangle angle ofof of 4545 with 45°° andand° and the rakerake smallrake −9−9 −955 relocated°°5.. °ThisThis. This solutionsolution solution (h = 4.5 isis km) issimilarsimilar similar focal toto depthto thethe the oneone ofone obtained theobtained obtained earthquake. forfor for thethe the However,secondsecond second earthquakeearthquakeearthquakethe average ofof of 44 centroid March4March March 20212021 depth 2021 asas ofas wellwell ~9well kmasas as toto falls to oneone one well ofof of thethe inside the twotwo two theaverageaverage average subsidence NPsNPs NPs foundfound domain.found fromfrom from variousvarious various seismicseismicseismic The faultfault fault best-ﬁt--planeplane-plane solutions geodeticsolutions solutions source(Table(Table (Table solution4).4). 4). TheThe The geodeticgeodetic forgeodetic the earthquakemomentmoment moment (Table(Table (Table of 12 5)5) March 5) resultsresults results 2021 inin in magni-magni- (Tablemagni-5, Figure 22) provides a seismic fault dipping to about SW with strike ESE–WNW (117◦), dip

Geosciences 2021, 11, 311 19 of 24 Geosciences 2021, 11, 311 19 of 24

angle of 45◦ and rake −95◦. This solution is similar to the one obtained for the second tude Mw = 5.7, whichearthquake is consistent of 4 March with 2021 the as well Mw~5.6 as to one determined of the two fromaverage seismic NPs found moment from various tensor solutions (Tableseismic S1). fault-plane Seismic solutions slip on (Table the4). fault The geodetic occurred moment mainly (Table in5 ) two results distinct in magnitude patches at depth ofM ~5w = km 5.7, the which first is consistentand ~10 km with the the second Mw~5.6 (Fi determinedgure 22). fromThe maximum seismic moment slip tensor of ~1.0 m, however,solutions occurred (Table within S1). Seismicthe deeper slip onpatch, the fault which occurred is situated mainly to in the two southwest distinct patches at depth of ~5 km the ﬁrst and ~10 km the second (Figure 22). The maximum slip of ~1.0 m, with respect to the however,first. occurred within the deeper patch, which is situated to the southwest with respect to the ﬁrst.

(a) (b)

FigureFigure 22. Modeled 22. Modeled geodetic geodetic seismic fault seismic for the fault earthquake for the ofearthquake 12 March 2021 of 12 in (Marcha) 3D illustration 2021 in ( anda) 3D (b )illustration vertical proﬁle alongand strike (b) (ESE–WNW). vertical profile along strike (ESE–WNW).

4. Discussion 4. Discussion The seismotectonic implications of the results obtained from the source solutions for The seismotectonicthe first implications and second earthquakes of the results are importantobtained sincefrom theythe source imply that solutions the fault for activated the first and secondwith earthquakes the second earthquake are important is subparallel since they and imply antithetic that to the the onefault associated activated with the with the second earthquakefirst earthquake is subparallel (Figures 23 and and 24 antithetic). This is similar to the to one the seismotectonicsassociated with characterizing the first earthquake (Figuresthe strong 23 earthquakeand 24). This sequence is similar that rupturedto the seismotectonics the area of the easterncharacterizing Gulf of Corinth on 24 February 1981 (M = 6.58), 25 February 1981 (M = 6.32) and again on 4 March the strong earthquake sequence that rupturedw the area of the easternw Gulf of Corinth on 1981 (Mw = 6.23) (magnitudes taken from [40]) in Central Greece to the south of Thessaly 24 February 1981 province(Mw = 6.58), (Figure 251 ).February During the 1981 first two(Mw 1981 = 6.32) earthquakes, and again surface on 4 fault-ruptures March 1981 striking (Mw = 6.23) (magnitudesroughly WSW–ENE taken from and [40]) dipping in Central to ~NNW Greece were detected, to the whilesouth a of roughly Thessaly parallel but province (Figure 1).antithetic During fault the first was activatedtwo 1981 with earthquakes, the third earthquake, surface fault e.g.,-rupture [41,42].s However, striking during roughly WSW–ENEMarch and 2021 dipping the faults to ~NNW activated were are detected, blind, unmapped while a and rough remainedly parallel unknown but so far. antithetic fault wasComplex activated normal with fault the systems third thatearthquake involve antithetic, e.g., [41,42]. fault segments However, have during been described in several cases of seismic sequences with multiple earthquake ruptures, e.g., the Kozani- March 2021 the faults activated are blind, unmapped and remained unknown so far. Grevena (NW Greece) earthquake (Mw = 6.5) of 13 May 1995 [43], the 26 December 2003 Complex normal fault systems that involve antithetic fault segments have been described Ban earthquake (Mw = 6.5) in Iran [44], and the Central Italy earthquakes on 24 August in several cases ofand seismic 26 and sequences 30 October with 2016 (Mmultiplew = 6.1, 5.9,earthquake and 6.5) [45 ruptures,]. e.g., the Koza- ni-Grevena (NW Greece) earthquake (Mw = 6.5) of 13 May 1995 [43], the 26 December 2003 Ban earthquake (Mw = 6.5) in Iran [44], and the Central Italy earthquakes on 24 August and 26 and 30 October 2016 (Mw = 6.1, 5.9, and 6.5) [45]. The four times less amplitude of the geodetic subsidence displacement, d, found for the second strong earthquake of 4 March (d = −10 cm), with respect to the one found for the first shock of 3 March (d = −40 cm), is better explained by a moment magnitude of 6.0 or 6.1, produced by national seismological institutes for the second earthquake (Table S1), or even by our geodetic estimation of Mw = 6.2, than by the Mw = 6.3 determined by GEOFON/GFZ, which perhaps is an overestimation. The third earthquake of 12 March ruptured to the NW continuation of the second rupture. The comparison of LoS displacements found from the observed and modeled InSAR data shows small residuals at the southern part of the deformation domain ob-

Geosciences 2021, 11, 311 20 of 24

tained from the ascending mode (Figure 20). However, the residuals are equal to zero at the descending mode (Figure 21). For this reason, the last LoS displacement is a tentatively preferred one and indicates subsidence of ~5 cm in the central side of the deformation field and uplift of ~2.5 cm in the NE side. Our best-fit geodetic source solution for the third earthquake (strike/dip/rake: 117°/45°/−95°) is quite similar to the one obtained for the second earthquake and consistent with published moment tensor solutions (average values for one nodal plane: 106°/39°/−105°). The mean of the dip angle determined geodetically (Table 5) and from moment tensor solutions (Table 4) is 42°. A Mw = 5.7 was determined, which is consistent with the Mw = 5.6 found from moment tensor solutions, while the maximum slip of 1.0 m occurred at depth of ~10 km. However, further research is needed in the future, e.g., regarding space–time aftershocks distribution, to verify that our preferred fault geometry is a valid one. Otherwise, we do not rule out that a fault geometry Geosciences 2021, 11, 311 20 of 24 similar to that associated with the first shock may fit better to the case of the 12 March aftershock.

FigureFigure 23. 23.Inferred Inferred surfacesurface projectionsprojections ofof thethe blind blind seismicseismic faultsfaults thatthat activated activated withwith thethe 3 3 (SF1, (SF1, ZarkosZarkos Fault), Fault), 4 (SF2, 4 (SF2, Domeniko Domeniko Fault) Fault) and and12 March 12 March 2021 (SF3, 2021 Kalivia (SF3, Kalivia Fault) earthquakes Fault) earthquakes according according to our interpretation of the combined seismic and geodetic analysis. Strikes and dip angles to our interpretation of the combined seismic and geodetic analysis. Strikes and dip angles of SF1, of SF1, SF2, SF3 were adopted as the respective averages of strikes and dip angles calculated from Geosciences 2021, 11, 311 SF2, SF3 were adopted as the respective averages of strikes and dip angles calculated from21 of the 24 the different methods (Tables 4 and 5):◦ 314◦ °/41° for SF1,◦ 123◦°/44° for SF2,◦ 112°◦/42° for SF3. Positions differentof the three methods faults (Tables were 4adopted and5): 314 by taking/41 for into SF1, account 123 /44 the forfault SF2, models, 112 /42 resultedfor SF3. from Positions the inver- of thesion three of teleseis faults weremic P adopted waveforms by taking and o intof InSAR account data, the as fault well models,as the respective resulted from relocated the inversion focal depth of teleseismicasperity(Table 3) thatand P waveforms the ruptured. average and dip A of different angle InSAR for data, patterneach as fault. well was asLengths the found respective of regardingabout relocated 23, 17 the and focal ~6 11-month depthkm were (Table lasting calcu-3) and the average dip angle for each fault. Lengths of about 23, 17 and 11 km were calculated for SF1, forlatedeshocks for SF1, preceding SF2 and SF3, the respectively, 25 October by 2018 taking mainshock into account (M rupturew = 6.8), lengths associated determined with from the SF2thrustour and seismic-oblique SF3, respectively,and- slipgeodetic rupture by fault taking offshoremodels into (Figures account Zakynthos rupture8, 11, Isl.,14 lengths and Ionian 22) determined and Sea a [47].worldwide from Namely, our empirical seismic the fore- andrela- tionship between subsurface rupture length and magnitude found for normal earthquakes [46]. geodeticshocks bounded fault models the (Figures mainshock8, 11, asperity14 and 22 up) and-dip a worldwideand at the empiricalnorth of it. relationship However, between the two subsurfaceΕpicenters rupture (stars) and length magnitudes and magnitude of the strong found forearthquakes normal earthquakes of 3, 4 and [1246 ].March Epicenters 2021, as (stars) well andas of largestthe main imminent faults (black foreshocks lines) in the (MThessalyw = 4 .1,area, M asw in= Figure4.8) occurred 1. very close to the 2018 magnitudesmainshock of hypocenter the strong earthquakes as happened of 3, also 4 and with 12 March the imminent 2021, as well foreshocks as of the main of the faults 3 March (black lines) in the Thessaly area, as in Figure1. 2021 shock.The relocated epicenters of both the second and third earthquakes fall at the NE margin of the down-dip projection of the geodetic fault solutions. One may argue that epicentral locations shifted by a few km to SW would fit better the SW-dipping fault planes. However, such an epicentral shift requires focal depths of ~14.5 and ~11.5 km instead of the very shallow relocated hypocenters of 4.0 and 4.5 km found for the 2nd and 3rd earthquakes, respectively. However, the surface projection of the average centroid depth of ~9–10 km of these earthquakes is shifted towards SW and falls well inside the subsidence domains. The 3 March earthquake was preceded by a short-living foreshock sequence occurring from 28 February to 3 March 2021. Most of the foreshocks occurred within the main

FigureFigure 24.24. Not-in-scaleNot-in-scale 3D3D cartooncartoon thatthat illustratesillustrates thethe threethree normalnormal seismicseismic faultfault (SF)(SF) segmentssegments thatthat activatedactivated withwith thethe earthquakesearthquakes ofof thethe 33 (SF1),(SF1), 44 (SF2)(SF2) andand 1212 March March (SF3) (SF3) 2021. 2021.

TheOur four seismotectonic times less amplitude interpretation of the for geodetic the March subsidence 2021 earthquake displacement, sequence d, found is dif-for theferent second from strong the one earthquake that considers of 4 March that the(d = three−10 cm), earthquakes with respect ruptured to the at one spatially found forse- thequential ﬁrst shock segments of 3 Marchof a normal (d = − fault40 cm), system is better striking explained SE–NW by and a moment dipping magnitude NE [16]. Such of 6.0 a orfault 6.1, model produced requires by national focal depths seismological of ~15 km institutes to explain for the the position second of earthquake the epicenters (Table of S1 the), second and third earthquakes at the northeast sides of the respective surface deformation fields, a point already discussed as regards our seismotectonic model. In addition, the NE/SW uplift/subsidence pattern recognized in our geodetic results for the second and third earthquakes (Figures 13, 16 and 21) is better explained by the fault model proposed in this paper. However, our interpretation is consistent with the suggestion [16] that the activated fault system is clearly belonging to the Tyrnavos Graben that started forming in the middle-late Pleistocene, and whose bordering structures are still in a growing phase.

5. Conclusions From our results we concluded that the sequence of strong earthquakes that occurred in Thessaly province, Central Greece, during March 2021 was produced by a quite complicated rupture process. The three strong earthquakes of the 3, 4 and 12 March 2021, were associated with three main blind, unmapped and unknown so far fault segments of nearly pure normal faulting with a small strike-slip component. The entire system of normal fault segments activated forms a graben-like structure. The first earthquake of 4 March (Mw = 6.3) was produced by a fault segment striking SE-NW and dipping towards NE. With the strong earthquake of 4 March (Mw = 6.2) the rupture propagated further NW but in an antithetic fault segment dipping ~SW and striking ~SE–NW. A segment of possibly the same fault, striking ESE–WNW was ruptured with the strong earthquake (Mw = 5.7) of 12 March. The new findings shed light in the Thessaly seismotectonics, particularly in the north side of this area, which remained seismically silent since AD 1781.

Geosciences 2021, 11, 311 21 of 24

or even by our geodetic estimation of Mw = 6.2, than by the Mw = 6.3 determined by GEOFON/GFZ, which perhaps is an overestimation. The third earthquake of 12 March ruptured to the NW continuation of the second rupture. The comparison of LoS displacements found from the observed and modeled InSAR data shows small residuals at the southern part of the deformation domain obtained from the ascending mode (Figure 20). However, the residuals are equal to zero at the descending mode (Figure 21). For this reason, the last LoS displacement is a tentatively preferred one and indicates subsidence of ~5 cm in the central side of the deformation field and uplift of ~2.5 cm in the NE side. Our best-fit geodetic source solution for the third earthquake (strike/dip/rake: 117◦/45◦/−95◦) is quite similar to the one obtained for the second earthquake and consistent with published moment tensor solutions (average values for one nodal plane: 106◦/39◦/−105◦). The mean of the dip angle determined ◦ geodetically (Table5) and from moment tensor solutions (Table4) is 42 .AMw = 5.7 was determined, which is consistent with the Mw = 5.6 found from moment tensor solutions, while the maximum slip of 1.0 m occurred at depth of ~10 km. However, further research is needed in the future, e.g., regarding space–time aftershocks distribution, to verify that our preferred fault geometry is a valid one. Otherwise, we do not rule out that a fault geometry similar to that associated with the first shock may fit better to the case of the 12 March aftershock. The relocated epicenters of both the second and third earthquakes fall at the NE margin of the down-dip projection of the geodetic fault solutions. One may argue that epicentral locations shifted by a few km to SW would fit better the SW-dipping fault planes. However, such an epicentral shift requires focal depths of ~14.5 and ~11.5 km instead of the very shallow relocated hypocenters of 4.0 and 4.5 km found for the 2nd and 3rd earthquakes, respectively. However, the surface projection of the average centroid depth of ~9–10 km of these earthquakes is shifted towards SW and falls well inside the subsidence domains. The 3 March earthquake was preceded by a short-living foreshock sequence occurring from 28 February to 3 March 2021. Most of the foreshocks occurred within the main asperity that ruptured. A different pattern was found regarding the ~6-month lasting foreshocks preceding the 25 October 2018 mainshock (Mw = 6.8), associated with the thrust-oblique- slip rupture offshore Zakynthos Isl., Ionian Sea [47]. Namely, the foreshocks bounded the mainshock asperity up-dip and at the north of it. However, the two largest imminent foreshocks (Mw = 4.1, Mw = 4.8) occurred very close to the 2018 mainshock hypocenter as happened also with the imminent foreshocks of the 3 March 2021 shock. Our seismotectonic interpretation for the March 2021 earthquake sequence is different from the one that considers that the three earthquakes ruptured at spatially sequential segments of a normal fault system striking SE–NW and dipping NE [16]. Such a fault model requires focal depths of ~15 km to explain the position of the epicenters of the second and third earthquakes at the northeast sides of the respective surface deformation fields, a point already discussed as regards our seismotectonic model. In addition, the NE/SW uplift/subsidence pattern recognized in our geodetic results for the second and third earthquakes (Figures 13, 16 and 21) is better explained by the fault model proposed in this paper. However, our interpretation is consistent with the suggestion [16] that the activated fault system is clearly belonging to the Tyrnavos Graben that started forming in the middle-late Pleistocene, and whose bordering structures are still in a growing phase.

5. Conclusions From our results we concluded that the sequence of strong earthquakes that occurred in Thessaly province, Central Greece, during March 2021 was produced by a quite complicated rupture process. The three strong earthquakes of the 3, 4 and 12 March 2021, were associated with three main blind, unmapped and unknown so far fault segments of nearly pure normal faulting with a small strike-slip component. The entire system of normal fault segments activated forms a graben-like structure. The ﬁrst earthquake of 4 March (Mw = 6.3) was produced by a fault segment striking SE-NW and dipping towards NE. Geosciences 2021, 11, 311 22 of 24

With the strong earthquake of 4 March (Mw = 6.2) the rupture propagated further NW but in an antithetic fault segment dipping ~SW and striking ~SE–NW. A segment of possibly the same fault, striking ESE–WNW was ruptured with the strong earthquake (Mw = 5.7) of 12 March. The new ﬁndings shed light in the Thessaly seismotectonics, particularly in the north side of this area, which remained seismically silent since AD 1781. The new knowledge acquired provides important insights for the better assessment of the seismic hazard level in the area.

Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/geosciences11080311/s1, Table S1: Source parameters of the three strongest earthquakes recorded in Thessaly area during March 2021; Table S2: Geometric features of the two nodal planes determined in moment tensor solutions produced by several seismological centers for the three strong earthquakes; Table S3: Specifications of the Sentinel-1 SAR SLC images used in our InSAR analysis; Table S4: Master and slave scenes forming interferometric pairs used to detect and measure deformation of the three earthquake events. Author Contributions: Conceptualization, E.L., G.A.P. and I.P.; methodology, G.A.P. and I.P.; software, A.A., A.K. and I.T.; data curation, A.A., A.K. and I.T.; writing—original draft preparation, G.A.P., A.A., A.K. and I.T.; writing—review and editing, all; supervision, G.A.P. All authors have read and agreed to the published version of the manuscript. Funding: This research received no external funding. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: https://scihub.copernicus.eu/, accessed on 7 June 2021. Acknowledgments: This paper is a contribution to the Post-Graduate Program “Environmental, Disaster, and Crises Management Strategies”, Department of Dynamic, Tectonic and Applied Geology, Faculty of Geology and Geoenvironment, National and Kapodistrian University of Athens, Greece. We are thankful to the three anonymous reviewers for their comments that contributed to improve the initial submission. Maps have been produced with the use of ArcMap 10.1 and Generic Mapping Tools-GMT 5 [48]. Conflicts of Interest: The authors declare no conflict of interest.

References 1. Lekkas, E.; Agorastos, K.; Mavroulis, S.; Kranis, H.; Skourtsos, E.; Carydis, P.; Gogou, M.; Katsetsiadou, K.N.; Papadopoulos, G.; Triantafyllou, I.; et al. The early March 2021 Thessaly (Greece) earthquake sequence. Newsl. Environ. Disaster Crises Manag. Strateg. 2021, 22, 1–195. [CrossRef] 2. Caputo, R. Geological and structural study of the recent and active brittle deformation of the Neogene-Quaternary Basins of Thessaly (Central Greece). Aristotle Univ. Thessalon. Sci. Ann. 1990, 2, 252. 3. Caputo, R. Morphotectonics and kinematics along the Tirnavos Fault, northern Larissa Plain, mainland Greece. Zeit. Für Geomorph. 1993, 94, 167–185. 4. Caputo, R.; Pavlides, S. Late Cainozoic geodynamical evolution of Thessaly and surroundings (central-northern Greece). Tectonophysics 1993, 223, 339–362. [CrossRef] 5. Caputo, R.; Helly, B.; Pavlides, S.; Papadopoulos, G. Palaeoseismological investigation of the Tyrnavos Fault (Thessaly, Central Greece). Tectonophysics 2004, 394, 1–20. [CrossRef] 6. Caputo, R.; Helly, B.; Pavlides, S.; Papadopoulos, G.A. Archaeo-and palaeoseismological investigations in Northern Thessaly (Greece): Insights for the seismic potential of the region. Nat. Hazards 2006, 39, 195–212. [CrossRef] 7. Tsodoulos, I.; Stamoulis, K.; Caputo, R.; Koukouvelas, I.; Chatzipetros, A.; Pavlides, S.; Gallousi, C.; Papachristodoulou, C.; Ioannides, K. Middle-Late Holocene earthquake history of the Gyrtoni Fault, Central Greece: Insight from optically stimulated luminescence (OSL) dating and paleoseismology. Tectonophysics 2016, 687, 14–27. [CrossRef] 8. Caputo, R.; Chatzipetros, A.; Pavlides, S.; Sboras, S. The Greek Database of Seismogenic Sources (GreDaSS): State-of-the-art for northern Greece. Ann. Geophys. 2013, 55, 859–894. [CrossRef] 9. Papadopoulos, G.A. Rupture zones of strong earthquakes in the Thessalia region, Central Greece. In Proceedings of the XXIII General Assembly, European Seismological Commission, Prague, Czech Republic, 7–12 September 1992; Geophysical Institute, Czechoslovak Academy of Science: Prague, Czech Republic, 1993; pp. 337–340. Geosciences 2021, 11, 311 23 of 24

10. Papadopoulos, G.A. Tectonic and seismic processes of various space and time scales in the Greek area. In Recent Evolution and Seismicity of the Mediterranean Region; Boschi, E., Mantovani, E., Morelli, A., Eds.; Kluwer: Dordrecht, The Netherlands, 1993; pp. 239–249. 11. Caputo, R. A possible seismic gap of Northern Thessaly, Greece, as inferred from geological data. Bull. Geol. Soc. Greece 1994, 30, 263–272. 12. Caputo, R. Inference of a Seismic Gap from Geological Data: Thessaly (Central Greece) as a Case Study. Ann. Geophys. 1995, 38, 1–19. Available online: https://www.annalsofgeophysics.eu/index.php/annals/article/view/4127 (accessed on 29 May 2021). [CrossRef] 13. Papadimitriou, E.; Karakostas, V. Episodic occurrence of strong (Mw≥6.2) earthquakes in Thessalia area (central Greece). Earth Planet Sci. Lett. 2003, 215, 395–409. [CrossRef] 14. Caputo, R. Comment on “Episodic occurrence of strong (Mw≥6.2) earthquakes in Thessalia area (central Greece)” by EE Pa- padimitriou and VG Karakostas [Earth Planet. Sci. Lett. 215 (2003) 395–409]. Earth Planet Sci. Lett. 2004, 231, 347–352. [CrossRef] 15. Papadimitriou, E.; Karakostas, V. Occurrence patterns of strong earthquakes in Thessalia area (Greece) determined by the stress evolutionary model. Earth Planet Sci. Lett. 2005, 235, 766–770. [CrossRef] 16. Tolomei, C.; Caputo, R.; Polcari, M.; Famiglietti, N.A.; Maggini, M.; Stramondo, S. The Use of Interferometric Synthetic Aperture Radar for Isolating the Contribution of Major Shocks: The Case of the March 2021 Thessaly, Greece, Seismic Sequence. Geosciences 2021, 11, 191. [CrossRef] 17. Pavlides, S.; Chatzipetros, A.; Sboras, S.; Kremastas, E.; Chatziioannou, A. The Northern Thessaly Strong Earthquakes of March 3 and 4 and Their Neotectonic Setting; Earthquake Geology Team, Aristotle University Thessaloniky: Thessaloniky, Greece, 2021; p. 10. Available online: http://eqgeogr.weebly.com/ (accessed on 7 June 2021). 18. Papadopoulos, G.A.; Lefkopoulos, A. Magnitude-distance relations for liquefaction in soil from earthquakes. Bull. Seism. Soc. Am. 1993, 83, 925–938. 19. Lomax, A.; Virieux, J.; Volant, P.; Berge-Thierry, C. Probabilistic earthquake location in 3D and layered models. In Advances in Seismic Event Location, 1st ed.; Thurber, C.H., Rabinowitz, N., Eds.; Springer: Dordrecht, The Netherlands, 2000; pp. 101–134. [CrossRef] 20. Lomax, A.A.; Michelini, A.; Curtis, A. Earthquake Location, Direct, Global-Search Methods. In Encyclopedia of Complexity and System Science, 2nd ed.; Meyers, R.A., Ed.; Springer: New York, NY, USA, 2014; pp. 1–33. [CrossRef] 21. Tarantola, A.; Valette, B. Generalized nonlinear inverse problems solved using the least squares criterion. Rev. Geophys. 1982, 20, 219–232. [CrossRef] 22. Karastathis, V.K.; Papoulia, J.; Di Fiore, B.; Makris, J.; Tsambas, A.; Stampolidis, A.; Papadopoulos, G.A. Deep structure investigations of the geothermal field of the North Euboean Gulf, Greece, using 3-D local earthquake tomography and Curie Point Depth analysis. J. Volcanol. Geotherm. Res. 2011, 206, 106–120. [CrossRef] 23. Karastathis, V.K.; Mouzakiotis, E.; Ganas, A.; Papadopoulos, G.A. High-precision relocation of seismic sequences above a dipping Moho: The case of the January-February 2014 seismic sequence on Cephalonia island (Greece). Solid Earth 2015, 6, 173–184. [CrossRef] 24. Hartzell, S.H.; Heaton, T.H. Inversion of strong ground motion and teleseismic waveform data for the fault rupture history of the 1979 Imperial Valley, California, earthquake. Bull. Seismol. Soc. Am. 1983, 73, 1553–1583. [CrossRef] 25. Hartzell, S.H.; Liu, P.; Mendoza, C. The 1994 Northridge, California earthquake: Investigation of rupture velocity, rise time and high-frequency radiation. J. Geophys. Res. 1996, 101, 20091–20108. [CrossRef] 26. Mendoza, C.; Hartzell, S.H. Finite-Fault Source Inversion Using Teleseismic P Waves: Simple Parameterization and Rapid Analysis. Bull. Seism. Soc. Am. 2013, 103, 834–844. [CrossRef] 27. Papadopoulos, G.A.; Agalos, A.; Charalampakis, M.; Kontoes, C.; Papoutsis, I.; Atzori, S.; Svigkas, S.; Triantafyllou, I. Fault models for the Bodrum-Kos tsunamigenic earthquake (Mw6.6) of 20 July 2017 in the east Aegean Sea. J. Geodyn. 2019, 131.[CrossRef] 28. Langston, C.A.; Helmberger, D.V. A procedure for modeling shallow dislocation sources. Geophys. J. Royal. Astron. Soc. 1975, 42, 117–130. [CrossRef] 29. Heaton, T. The 1971 San Fernando earth quake: A double event? Bull. Seism. Soc. Am. 1982, 72, 2037–2062. [CrossRef] 30. Bamler, R.; Hartl, P. Synthetic aperture radar interferometry. In Inverse Problems; IOP Publishing Limited: Bristol, UK, 1998. 31. Massonnet, D.; Rossi, M.; Carmona, C.; Adragna, F.; Peltzer, G.; Feigl, K.; Rabaute, T. The displacement field of the Landers earthquake mapped by radar interferometry. Nature 1993, 364, 138–142. [CrossRef] 32. Zebker, H.A.; Rosen, P.A.; Goldstein, R.M.; Gabriel, A.; Werner, C.L. On the derivation of coseismic displacement fields using differential radar interferometry: The Landers earthquake. J. Geophys. Res. 1994, 99, 19617–19634. [CrossRef] 33. Massonnet, D.; Feigl, K.L. Radar interferometry and its application to changes in the earth’s surface. Rev. Geophys. 1998, 36, 441–500. [CrossRef] 34. Bally, P. Satellite Earth Observation for Geohazard Risk Management—The Santorini Conference, Santorini, Greece, 21–23 May 2012; ESA Publ. STM-282; European Space Agency: Paris, France, 2014. [CrossRef] 35. Wright, T.J.; Parsons, B.E.; Lu Toward, Z. Mapping surface deformation in three dimensions using InSAR. Geophys. Res. Lett. 2004, 31, L01607. [CrossRef] 36. Raucoules, D.; De Michele, M.; Malet, J.-P.; Ulrich, P. Time-variable 3D ground displacements from high-resolution synthetic aperture radar (SAR). Application to La Valette landslide (South French Alps). Remote Sens. Environ. 2013, 139, 198–204. [CrossRef] 37. SARmapModeling Tutorial 2020, Version 5.5.3. p. 24. Available online: https://www.sarmap.ch/tutorials/ModelingTutorial_v5 53.pdf (accessed on 27 March 2021). Geosciences 2021, 11, 311 24 of 24

38. Okada, Y. Internal deformation due to shear and tensile faults in a half-space. Bull. Seism. Soc. Am. 1992, 82, 1018–1040. 39. Costantini, M. A novel phase unwrapping method based on network programming. IEEE Trans. Geosci. Remote Sens. 1998, 36, 813–821. [CrossRef] 40. International Seismological Centre. ISC-GEM Earthquake Catalogue; ISC: Berkshire, UK, 2021. [CrossRef] 41. Jackson, J.A.; Gagnepain, J.; Houseman, G.; King, G.C.P.; Papadimitriou, P.; Souﬂeris, C.; Virieux, J. Seismicity, normal faulting and the geomorphological development of the Gulf of Corinth (Greece): The Corinth earthquakes of February and March 1981. Earth Planet. Sci. Lett. 1982, 57, 377–397. [CrossRef] 42. Hubert, A.; King, G.C.P.; Mayer, B.; Papanastasiou, D. Fault re-activation, stress interaction and rupture propagation of the 1981 Corinth earthquake sequence. Earth Planet Sci. Lett. 1996, 142, 573–585. [CrossRef] 43. Resor, P.G.; Pollard, D.D.; Wright, T.J.; Beroza, G.C. Integrating high-precision aftershock locations and geodetic observations to model coseismic deformation associated with the 1995 Kozani-Grevena earthquake, Greece. J. Geophys. Res. 2005, 110, B09402. [CrossRef] 44. Parcharidis, I.; Zare, M.; Foumelis, M.; Lagios, E. Seismotectonic investigation on the Bam earthquake prone area (Iran) based on ASAR interferometry. In Proceedings of the 2nd International Conference Recent Advances Space Technologies, Istanbul, Turkey, 9–11 June 2005; pp. 673–677. [CrossRef] 45. Cheloni, D.; De Novellis, V.; Albano, M.; Antonioli, A.; Anzidei, M.; Atzori, S.; Avallone, A.; Bignami, C.; Bonano, M.; Calcaterra, S. Geodetic model of the 2016 Central Italy earthquake sequence inferred from InSAR and GPS data. Geophys. Res. Lett. 2017, 44, 6778–6787. [CrossRef] 46. Wells, D.L.; Coppersmith, K.J. New empirical relationships among magnitude, rupture length, rupture width, rupture area, and surface displacement. Bull. Seism. Soc. Am. 1994, 84, 974–1002. 47. Papadopoulos, G.A.; Agalos, A.; Minadakis, G.; Triantafyllou, I.; Krassakis, P. Short-Term Foreshocks as Key Information for Mainshock Timing and Rupture: The Mw6.8 25 October 2018 Zakynthos Earthquake, Hellenic Subduction Zone. Sensors 2020, 20, 5681. [CrossRef][PubMed] 48. Wessel, P.; Smith, W.H.F.; Scharroo, R.; Luis, J.; Wobbe, F. Generic Mapping Tools: Improved Version Released. EOS Trans. AGU 2013, 94, 409–410. [CrossRef]