Croatia) Earthquake Inferred from the Dinsar, Seismological, and Geological Observations

remote sensing

Article Constraints on Complex Faulting during the 1996 Ston–Slano (Croatia) Earthquake Inferred from the DInSAR, Seismological, and Geological Observations

Marin Govorˇcin 1 , Marijan Herak 2, Bojan Matoš 3,*, Boško Pribiˇcevi´c 1 and Igor Vlahovi´c 3

1 Department of Geomatics, Faculty of Geodesy, University of Zagreb, Kaˇci´ceva26, HR-10000 Zagreb, Croatia; [email protected] (M.G.); [email protected] (B.P.) 2 Department of Geophysics, Faculty of Science, University of Zagreb, Horvatovac 95, HR-10000 Zagreb, Croatia; [email protected] 3 Department of Geology and Geological Engineering, Geology and Petroleum Engineering, Faculty of Mining, University of Zagreb, Pierottijeva 6, HR-10000 Zagreb, Croatia; [email protected] * Correspondence: [email protected]; Tel.: +385-01-553-5787

Received: 6 March 2020; Accepted: 1 April 2020; Published: 4 April 2020

Abstract: This study, involving remote sensing, seismology, and geology, revealed complex faulting during the mainshock of the Ston–Slano earthquake sequence (5 September, 1996, Mw = 6.0). The observed DInSAR interferogram fringe patterns could not be explained by a single fault rupture. Geological investigations assigned most of the interferogram features either to previously known faults or to those newly determined by ﬁeld studies. Relocation of hypocentres and reassessment of fault mechanisms provided additional constraints on the evolution of stress release during this sequence. Available data support the scenario that the mainshock started with a reverse rupture with a left-lateral component on the Slano fault 4.5 km ESE of Slano, at the depth of about 11 km. The rupture proceeded unilaterally to the NW with the velocity of about 1.5 km/s for about 11 km, where the maximum stress release occurred. DInSAR interferograms suggest that several faults were activated in the process. The rupture terminated about 20 km away from the epicentre, close to the town of Ston, where the maximum DInSAR ground displacement reached 38 cm. Such a complicated and multiple rupture has never before been documented in the Dinarides. If this proves to be a common occurrence, it can pose problems in deﬁning realistic hazard scenarios, especially in deterministic hazard assessment.

Keywords: southern Dinarides; Ston–Slano earthquake; earthquake relocation; DInSAR; complex faulting; coseismic deformation analysis

1. Introduction

The Ston–Slano earthquake sequence, with the mainshock of 5 September, 1996 (Mw = 6.0, Imax = VIII Medvedev–Sponheuer–Karnik scale, henceforth MSK), is the most important and the largest one in the southern Dalmatia zone, i.e., Dubrovnik epicentral area that occurred since the catastrophic Dubrovnik earthquake of 1667 (epicentral intensity Io = IX MSK). Described in detail by Markušićet al. [1] and Herak et al. [2], the mainshock was felt about 400 km away, and caused devastation at several localities in the epicentral area, where about 1400 buildings were damaged, of which 474 became uninhabitable [3]. Especially affected was the municipality of Ston. The old city centre suffered the most—wide cracks appeared in bearing walls (up to 10 cm wide) that also bulged and/or tilted, roofs and gables collapsed, etc. (Figure1). Peak horizontal ground acceleration of 0.64 g recorded in Ston [4] is the largest ever observed in Croatia.

Remote Sens. 2020, 12, 1157; doi:10.3390/rs12071157 www.mdpi.com/journal/remotesensing RemoteRemote Sens. Sens. 20202020, 12, 12, x, FOR 1157 PEER REVIEW 2 of2 23of 22

44 Figure 1. Damage in Ston ((a) Municipality building, (b) typical stone masonry building, (c) St. Blasius 45 Figurechurch) 1. Damage and in the in village Ston (( ofa) Mravinca Municipality (d), caused building by the, (b mainshock) typical stone on 5 masonry September, building 1996 (photographed, (c) St. Blasius 46 churchby the) andﬁeld crewin the of thevillage Department of Mravinca of Geophysics, (d), caused Zagreb, by during the mainshock macroseismic on survey, 5 September September, 1996 47 (pho1996)tographed [5]. by the field crew of the Department of Geophysics, Zagreb, during macroseismic 48 survey, September 1996) [5]. The earthquake sequence lasted for about a year, with more than 1800 located aftershocks within 49 50The km fromearthquake the mainshock’s sequence epicentre. lasted for Numerical about a year, modeling with more of the than aftershocks’ 1800 located rate ofaftershocks occurrence within was 50 50 donekm from by Herak the mainshock’s et al. [2]. The epicentre. largest aftershocks Numerical occurred modeling on 9 September,of the aftershocks’ 1996 (15:57 rate UTC, of Moccurrencew = 5.3, I = VII MSK) and 17 September, 1996 (13:45 UTC, M = 5.4). 51 wasmax done by Herak et al. [2]. The largest aftershocks occurredw on 9 September, 1996 (15:57 UTC, Mw The Ston–Slano earthquake occurred within the southern part of the Dinarides, c. 600 km long 52 = 5.3, Imax = VII MSK) and 17 September, 1996 (13:45 UTC, Mw = 5.4). fold-and-thrust orogenic belt structurally positioned along the NE margin of the Adriatic Sea, i.e., NE 53 The Ston–Slano earthquake occurred within the southern part of the Dinarides, c. 600 km long part of the Adria Microplate (Figure3). NW–SE striking Dinarides were uplifted during the Late Eocene 54 fold-and-thrust orogenic belt structurally positioned along the NE margin of the Adriatic Sea, i.e., NE to Oligocene due to the Adria Microplate–European Plate collision, where Adria Microplate acted as 55 part of the Adria Microplate (Figure 2). NW–SE striking Dinarides were uplifted during the Late a rigid indenter ([6,7] with references therein) in respect to the European foreland (i.e., Tisza–Dacia 56 Eocene to Oligocene due to the Adria Microplate–European Plate collision, where Adria Microplate Mega Unit; Figure3). Southwest of the Sava Suture Zone (Figure3), Internal Dinarides are composed 57 acted as a rigid indenter ([6,7] with references therein) in respect to the European foreland (i.e., Tisza– of several nappe systems (Figure3), i.e., Neotethyan ophiolitic units (Figure3), Flysch units, and thrust 58 Daciasheets Mega composed Unit; Figure of oceanic 2). So depositsuthwest ([ 6of,7] the with Sava references Suture therein). Zone (Figure The proximal 2), Internal part ofDinarides the Adria are 59 composedmargin is of represented several nappe by External systems Dinaridic (Figure units 2), (Dalmatiani.e., Neotethyan Zone andophiolitic High Karst units Unit; (Figure Figure 2),3), Flysch i.e., 60 unitsunits, and mostly thrust deposited sheets composed on the Adriatic of oceanic Carbonate deposits Platform ([6,7] (see with [8,9 references] and references therein). therein). The proximal 61 part of Thethe Adria epicentral margin area is of represented the Ston–Slano by sequenceExternal (FigureDinaridic3) occurred units (Dalmatian within the southernZone and Dalmatian High Karst 62 Unit;Zone Figure of the 2), External i.e., units Dinarides mostly (Figures deposited2 and 3 on), very the Adriaticclose to the Carbonate tectonic contact Platform with (see the [8 High,9] and 63 referencesKarst Nappe therein). towards NE (Figure2). Dalmatian zone is composed of folded and faulted Upper 64 Jurassic–CretaceousThe epicentral area carbonates of the depositedSton–Slano on sequence the Adriatic (Figure Carbonate 2) occurred Platform within ([8] and the references southern 65 Dalmatiantherein). Zone These of carbonates the External are covered Dinarides by Palaeogene(Figures 2 limestonesand 3), very and close ﬂysch to depositsthe tectonic (approximately contact with 66 the3800 High m thickKarst succession Nappe towards [10]). The NE High (Figure Karst 3). Nappe Dalmatian Unit in the zone hanging is composed wall of the of NE folded inclined and reverse faulted 67 Upperfault isJurassic composed–Cretaceous of c. 5000 carbonates m thick Mesozoic deposited to Palaeogene on the Adriatic succession Carbonate (Figure2;[ Platform10,11]). External ([8] and 68 references therein). These carbonates are covered by Palaeogene limestones and flysch deposits 69 (approximately 3800 m thick succession [10]). The High Karst Nappe Unit in the hanging wall of the 70 NE inclined reverse fault is composed of c. 5000 m thick Mesozoic to Palaeogene succession (Figure 71 3; [10,11]). External Dinarides are intensely folded and faulted as a result of intensive convergence of

Remote Sens. 2020, 12, 1157 3 of 23

Dinarides are intensely folded and faulted as a result of intensive convergence of the Adria Microplate and the European Plate. Still ongoing convergence, with rates of up to 4.17 mm/yr (see [12,13] for details), is evidenced by recent seismicity in the area that accommodate stress within the collisional zone of the undeformed part of the Adriatic Microplate and the External–Internal Dinarides transition Remote Sens.zone 2020 [14, 12,15, x]. FOR PEER REVIEW 4 of 22

90 Figure 3.Figure Geological 2. Geological map mapof the of theSton Ston–Slano–Slano area area (simpliﬁed(simplified after after [10 ,[1011]),11]) at the at NE the margin NE margin of the of the Dalmatian Zone close to the contact with the High Karst Nappe of the External Dinarides. Legend: 91 Dalmatian Zone close to the contact with the High Karst Nappe of the External Dinarides. Legend: HR: Croatia; B-H: Bosnia and Herzegovina; T3: Upper Triassic (predominantly dolomites); J1,J2, 3 1 2 3 92 HR: Croatia;J3: Lower, B-H: Middle,Bosnia andand Upper Herzegovina; Jurassic (limestones T : Upper with Triassic dolomites (predominantly in the uppermost dolomites); part); K1, J , J , J : 93 Lower, MiddleK1,2,K2:, Lower,and Upper Lower /JurassicUpper transition, (limestones and Upper with Cretaceousdolomites (predominantly in the uppermost limestones, part); some K1, K1,2, K2: 94 Lower, Lower/Upperdolomites in the transition middle part);, and Pc, Upp E, Eer1,2 Cretaceous: Palaeocene (predominantly and Lower–Middle limestones, Eocene (foraminifera some dolomites limestones); E2,3: Middle–Upper Eocene (marls and ﬂysch, i.e., alternation of marls and carbonate 95 in the middle part); Pc, E, E1,2: Palaeocene and Lower–Middle Eocene (foraminifera limestones); E2,3: sandstones); Q: Quaternary (silts, sands, and sandstones along the coast, alluvial sandstones and gravel 96 Middle–inUpper the hinterland). Eocene (marls and flysch, i.e., alternation of marls and carbonate sandstones); Q: 97 Quaternary (silts, sands, and sandstones along the coast, alluvial sandstones and gravel in the 98 hinterland).

99 Seismogenic sources in the southern Dalmatia zone, i.e., Dubrovnik epicentral area are 100 predominantly NW-striking thrust faults (maximum horizontal compressional stress—SHmax is NE– 101 SW trending compression, see Figures 2 and 3) with earthquakes confined to shallow crustal levels 102 (≤ 20 km in depth [18,19]). In addition, the historical seismicity (the Croatian Earthquake Catalogue 103 [20], updated in 2019, lists seven events exceeding estimated magnitude 6.0 in the circle of 50 km 104 radius around Dubrovnik since the 17th century), thousands of instrumentally recorded earthquakes 105 indicate still ongoing intense tectonic activity along the southern Dalmatian coastline within domains 106 of the mapped faults. 107 Considering the fact that the study area of the southernmost Dalmatia is characterised by the 108 highest seismic hazard in Croatia (area with several UNESCO Heritage Sites; e.g., Old City of 109 Dubrovnik), and that the complexity and properties of seismogenic faults in the area are still not 110 known well enough, the research objectives were to identify, and if possible, characterise source(s) of 111 coseismic deformation related to the 1996 Ston–Slano Mw = 6.0 earthquake. This task is directly linked 112 to the compilation process of the new seismic hazard map of Croatia, as fault source model is to be 113 considered as a branch in the hazard assessment logic tree structure. In order to provide insight and 114 effectively reinterpret local seismogenic architecture, we applied multidisciplinary approach 115 including DInSAR to analyse the coseismic deformation in correlation with additional seismological 116 data analysis and field structural-geological observations. Based on new findings which included 117 relocation of macroseismic and microseismic earthquake epicentres and (re)evaluation of the Fault 118 Mechanism Solutions (FMS), the observed DInSAR coseismic deformation, and defined structural 119 framework of the fault system in the area, we propose principal seismogenic sources and activation 120 scenario during the 1996 Ston–Slano earthquake sequence.

Remote Sens. 2020, 12, x FOR PEER REVIEW 3 of 22

72 the Adria Microplate and the European Plate. Still ongoing convergence, with rates of up to 4.17 73 mm/yr (see [12,13] for details), is evidenced by recent seismicity in the area that accommodate stress Remote Sens. 2020, 12, 1157 4 of 23 74 within the collisional zone of the undeformed part of the Adriatic Microplate and the External– 75 Internal Dinarides transition zone [14,15].

77 FigureFigure 2. 3.(a)( aTectonic) Tectonic map map of of the the Dinarides Dinarides and surroundingsurrounding areasareas with with major major tectonostratigraphic tectonostratigraphic 78 unitsunits and and fault fault systems systems (after (after [ [6,69],9] with with references). TheThe maximummaximum horizontal horizontal compressional compressional stress stress Hmax 79 (S (SHmax) orientations) orientations and and stress stress regimes regimes are are based based on on the the World World Stress MapMap datadata (see (see [ 16[16]] for for details). details). 80 YellowYellow arrows arrows represent represent calculated calculated GPS-derivedGPS-derived velocities velocities of of the the Adria Adria Microplate Microplate in mm in mm/yr/yr relative relative to 81 to Eurasia(after (after [ 12[12]] with with references). references). Tectonic Tectonic map map indicates indicates locations locations of seismic of seismic stations, stations, location location of 1996 of 82 1996Ston–Slano Ston–Slano Mw M6.0w earthquake,6.0 earthquake and, spatialand spatial extent extent of Figure of Figure2 (blue 3 rectangle). (blue rectangle). ( b) Tectonic (b) Tectonic sketch of sketch the 83 of Alpine–Carpathianthe Alpine–Carpathian and circum-Adriaticand circum-Adriatic orogenic orogenic system system with present-day with present deformation-day deformation fronts (e.g., fronts 84 (e.majorg., major thrust thrust and strike-slip and strike fault-slip systems). fault systems). Adria lithosphere Adria lithosphere (beige) is subducting (beige) is subducting beneath European beneath 85 Europeanslab (darker slab grey), (darker whereas grey), slab whereas motion slab vectors motion (purple vectors arrows) (purple indicate arrows) mantle indicate flow (simplified mantle flow 86 (simplifiedafter [17] with after references). [17] with Abbreviations: references). Abbreviations: AM: Adria Microplate; AM: Adria ALP: Microplate; Alps; APN: Apennines;ALP: Alps; DIN: APN: 87 Apennines;Dinarides; DIN HEL:: Dinarides; Hellenides. HEL: Hellenides. Seismogenic sources in the southern Dalmatia zone, i.e., Dubrovnik epicentral area are 88 predominantly NW-striking thrust faults (maximum horizontal compressional stress—SHmax is NE–SW trending compression, see Figures2 and3) with earthquakes confined to shallow crustal levels ( 20 km ≤ in depth [18,19]). In addition, the historical seismicity (the Croatian Earthquake Catalogue [20], updated in 2019, lists seven events exceeding estimated magnitude 6.0 in the circle of 50 km radius around Dubrovnik since the 17th century), thousands of instrumentally recorded earthquakes indicate

Remote Sens. 2020, 12, 1157 5 of 23 still ongoing intense tectonic activity along the southern Dalmatian coastline within domains of the mapped faults. Considering the fact that the study area of the southernmost Dalmatia is characterised by the highest seismic hazard in Croatia (area with several UNESCO Heritage Sites; e.g., Old City of Dubrovnik), and that the complexity and properties of seismogenic faults in the area are still not known well enough, the research objectives were to identify, and if possible, characterise source(s) of coseismic deformation related to the 1996 Ston–Slano Mw = 6.0 earthquake. This task is directly linked to the compilation process of the new seismic hazard map of Croatia, as fault source model is to be considered as a branch in the hazard assessment logic tree structure. In order to provide insight and effectively reinterpret local seismogenic architecture, we applied multidisciplinary approach including DInSAR to analyse the coseismic deformation in correlation with additional seismological data analysis and field structural-geological observations. Based on new findings which included relocation of macroseismic and microseismic earthquake epicentres and (re)evaluation of the Fault Mechanism Solutions (FMS), the observed DInSAR coseismic deformation, and defined structural framework of the fault system in the area, we propose principal seismogenic sources and activation scenario during the 1996 Ston–Slano earthquake sequence.

2. Seismological Observations The seismological part of the study comprised the relocation of earthquakes in the greater Ston–Slano area, (re)evaluation of Fault Mechanism Solutions (FMS) for the mainshock and the strongest aftershocks, and inversion of the parameters of the macroseismic ﬁeld for the mainshock of 5 September, 1996. The dataset used consisted of:

138,520 observed arrival times of seismic phases for 10,897 earthquakes that occurred between 1 • January, 1995 and 5 September, 2018 in the area within the radius of 90 km from the mainshock of the Ston–Slano sequence, based on the seismic data of the Department of Geophysics, Faculty of Science, University of Zagreb (DGFSUZ), and supplemented by arrival times published in the ISC bulletin [21] or in available seismic bulletins from the neighbouring countries. In addition, we have also repicked some onset times from the available seismograms from the analogue and digital seismogram archive of the DGFSUZ, as well as from the international digital seismogram archives [22,23]. A database of FMS and the corresponding first-motion polarity readings for earthquakes in Croatia • and the neighbouring regions since 1910 maintained by the DGFSUZ. Those were either manually read (for all digital seismograms, and for the analogue records obtained at stations of the Croatian Seismograph Network) or adopted from various bulletins and databases. All FMS related to the studied region were rechecked and recomputed using recent crustal models. Intensities observed due to the Ston–Slano mainshock in Croatia and Bosnia and Herzegovina • from the macroseismic archive of the DGFSUZ. The data on earthquake effects were collected by fieldwork, interviews, questionnaires sent into the greater epicentral area, and from other available sources (e.g., newspaper articles, official damage reports, etc.).

2.1. Earthquake Relocation At the time of the earthquake, the seismic network of Croatia was rather sparse and equipped with analogue instruments. The two closest stations were in Dubrovnik and Hvar, about 20 and 130 km away, respectively (Figure3). In order to better record the aftershock activity, a seismic station was deployed in Potomje on the Pelješac peninsula two days after the mainshock (Figure3). The instrumental locations of foci for this earthquake series as found in the Croatian Earthquake Catalogue ([20], last updated in 2019; see also [1]) were determined by assuming a simple three-layered model of the crust and the uppermost mantle with horizontal interfaces. Clearly, this average model is inadequate to describe the travel times of seismic waves in a region as complex as the studied one (see Remote Sens. 2020, 12, 1157 6 of 23

Sections1 and4), where Moho topography, as well as the lithology vary rapidly with the propagation azimuth and distance (e.g., [24,25]). We have, therefore, relocated all events in the 1995–2018 period using the Hyposearch program [26] that was recently upgraded with the option of implementing the Source-Specific Station Corrections (SSSC) [27]. The method consists of iterative locations using arrival time data corrected for the average observed residual for each specific station–phase–source triplet in the previous iteration. Such path-dependent station corrections are often used in regions of complex tectonics and geology (as is the case here) where 1D models are inappropriate to compute realistic arrival times in forward modeling, and 3D structural models do not exist (e.g., [28,29]). Each onset time used in the location was weighted depending on (i) the residual with respect to the theoretical travel-time, (ii) the wave type and the phase, (iii) the distance of the corresponding station, and (iv) the density of azimuth coverage around the station in question. The final locations of earthquake foci were defined as the median of 20 sets of locations obtained using different models of the crust and uppermost mantle, and different configuration parameters defining computation of SSSC and weighting. The final catalogue holds basic information for 10,897 events. Figure4 presents a subset of the most reliably located earthquakes. As can be seen from the figure, the epicentral area that was activated in the 1996–1997 sequence (Figure4a), is virtually aseismic ever since (Figure4b).

2.2. Fault Mechanism Solutions Fault Mechanism Solutions (FMS) were determined by a grid search of fault parameters (strike, dip, rake) that best fit the observed pattern of the first P-wave motion polarity and its amplitude. In addition, we have also consulted available Centroid Moment Tensor (CMT) solutions from Istituto Nazionale di Geofisica e Volcanologia (INGV, Italy), the National Earthquake Information Centre (NEIC, USA), the International Seismological Centre (ISC, UK), and the Global CMT Catalogue [30,31]. Table1 lists occurrence times, locations, magnitudes, and faulting parameters for the six largest events for which either a focal mechanism had been published before and/or the fault plane solution was computed in the present study. As can also be seen in Figure5, the epicentres reported by di fferent agencies, as well as the Best Double-Couple (BDC) of the CMT solutions or the FMS differ considerably. However, most of the solutions indicate predominance of dip-slip faulting, with some of them preferring a transpressive mechanism. This agrees well with the tectonic stress regime in this part of the External Dinarides, dominated by NE–SW oriented stress axis (Figures3 and5b). Remote Sens. 2020, 12, 1157 7 of 23 Remote Sens. 2020, 12, x FOR PEER REVIEW 6 of 22

169

170 Figure 4.4. ((aa)) TheThe mostmost reliably reliably located located epicentres epicentres in in a yeara year following following the the Ston–Slano Ston–Slano mainshock. mainshock. Only Only 171 locationslocations characterisedcharacterised by by a a standard standard error error of of the the epicentre epicentrσe H5 ≤ km5 km and and with with azimuthal azimuthal coverage coverage H ≤ 172 characterised by by a a gap gapγ  ≤150 150°◦ are are shown. shown. Symbols Symbols are are sized sized by by the the earthquake earthquake magnitude, magnitude, and and focal focal ≤ 173 depths areare indicated indicated by by the the symbol symbol colour colour according according to the to colourthe colour bar on bar the on right. the right. (b) The (b same) The as same (a) as 174 but(a) but for thefor timethe time period period of 21 of years, 21 years, starting starting a year a afteryear theafter mainshock the mainshock of 5 September, of 5 September 1996. , 1996.

175 2.2. Fault Mechanism Solutions 176 Fault Mechanism Solutions (FMS) were determined by a grid search of fault parameters (strike, 177 dip, rake) that best fit the observed pattern of the first P-wave motion polarity and its amplitude. In 178 addition, we have also consulted available Centroid Moment Tensor (CMT) solutions from Istituto 179 Nazionale di Geofisica e Volcanologia (INGV, Italy), the National Earthquake Information Centre 180 (NEIC, USA), the International Seismological Centre (ISC, UK), and the Global CMT Catalogue 181 [30,31]. Table 1 lists occurrence times, locations, magnitudes, and faulting parameters for the six 182 largest events for which either a focal mechanism had been published before and/or the fault plane 183 solution was computed in the present study. As can also be seen in Figure 5, the epicentres reported 184 by different agencies, as well as the Best Double-Couple (BDC) of the CMT solutions or the FMS differ

Remote Sens. 2020, 12, 1157 8 of 23

Table 1. Time of occurrence, hypocentral coordinates in decimal degrees ( st. errors in km), magnitudes, and the best double-couple solutions for the mainshock and ± ﬁve aftershocks, as obtained by various agencies and in this study (bold). φ, δ, λ are the strike, dip, and rake of the best double-couple (BDC) of the CMT solution or of the FMS. 1 DT: Centroid time minus hypocentre time; C: Centroid location; E: Microseismic epicentre; BDC: Best Double-Couple; FMS: The ﬁrst P-wave polarity Focal Mechanism Solution. 2 USGS [32], GCMT [33], ISC [34], ISC-GEM [35], RCMT [36].

1 1 φ1◦, δ1◦, λ1◦ φ2◦, δ2◦, 2 No. Date Time, DT (s) Lat., ◦N Lon., ◦E C, E Depth, km Mag. 1 Source λ2◦ (BDC or FMS) Mainshock 5 M 6.0, mb 328, 32, 92 1 20:44:17.3 DT = 8.1 s 42.78 17.77 C 15.0 w GCMT September, 1996 5.6, MS 6.0 146, 58, 89 (BDC) 285, 55, 30 2 20:44:09 42.80 17.94 E 10.0 M 6.0 USGS wc 177, 66, 141 (FMS) 345, 40, 102 3 20:44:11.3 42.78 17.92 E 13.3 M 5.9 ISC w 150, 52, 80 (FMS) M 6.0, mb 312, 69, 82 4 20:44:11.3 42.78 17.92 E 13.3 w ISC 5.6, MS 6.0 153, 23, 110 (FMS)

5 20:44:11 42.75 17.96 E 12.5 Mw 6.0 – ISCGEM 293, 53, 47 6 20:44:07.9 42.77 5 km 17.94 5 km E 11.4 5 ML 6.0 This study ± ± ± 170, 54, 132 (FMS) Aftershock 5 7 21:43:31.1 42.83 17.84 E 10.0 m 4.9 – RCMT (from PDE) September, 1996 b 21:43:34.6 321, 66 100 8 42.74 17.89 C 15.0 M 4.6 RCMT (from PDE) DT = 3.5 s w 119, 26, 70 (BDC)

9 21:43:30.3 42.77 4 km 17.85 4 km E 10.8 2 ML 4.9 – This study ± ± ± Aftershock 7 301, 59, 77 10 05:45:32.4 42.85 17.84 E 8.1 M 4.5 This study September, 1996 L 145,33, 111 (FMS) Aftershock 9 15:57:08.7 M 5.3, mb 301, 13, 58 11 43.03 17.55 C 15.0 w GCMT September, 1996 DT = 3.6 s 4.8, MS 5.0 154, 79, 97 (BDC)

12 15:57:05 42.77 17.873 E 10.0 Mwc 5.3 – USGS 223, 69, 1 13 15:57:04.6 42.75 2 km 17.83 2 km E 10.5 2 ML 5.0 This study ± ± ± 133, 89, 159 (FMS) Aftershock 17 13:45:27.9 M 5.4, mb 295, 35, 76 14 42.59 17.53 C 15.0 w GCMT September, 1996 DT = 5.1 s 5.4, MS 5.1 132, 56, 99 (BDC)

15 13:45:22 42.87 17.82 E 10.0 Mwc 5.4 – USGS 299, 33, 77 16 13:45:22.1 42.83 2 km 17.92 2 km E 12.8 1 ML 5.1 This study ± ± ± 134, 58, 98 (FMS) Aftershock 20 October, 319, 47, 83 17 15:00:01.7 42.78 1 km 17.93 1 km E 8.7 2 ML 4.6 This study 1996 ± ± ± 149, 43, 97 (FMS) Remote Sens. 2020, 12, x FOR PEER REVIEW 7 of 22

185 considerably. However, most of the solutions indicate predominance of dip-slip faulting, with some Remote Sens. 2020, 12, 1157 9 of 23 186 of them preferring a transpressive mechanism. This agrees well with the tectonic stress regime in this 187 part of the External Dinarides, dominated by NE–SW oriented stress axis (Figures 2 and 5b).

188 189 Figure 5. 5.( a)(a Epicentres) Epicentres and fault-plane and fault-plane solutions solutions (lower hemisphere (lower hemisphere projections, projections, coloured compressive coloured 190 quadrants)compressive reported quadrants) in Table reported1. The in same Table colour 1. The refers same to colour the same refers event. to the The same numbers event. refer The to numbers entries 191 inrefer Table to 1entries. White in and Table grey 1. dilatational White and quadrants grey dilatational denote focalquadrants mechanism denote solutions focal mechanism (FMS) and solution the bests 192 double-couple(FMS) and the best (BDC) double of the-couple centroid (BDC) moment of thetensor centroid (CMT) moment solutions, tensorrespectively (CMT) solutions, (see Table respectively1). HR: 193 Croatia;(see Table B-H: 1). Bosnia HR: Croatia; and Herzegovina. B-H: Bosnia (b) All and FMS Herzegovina. from the database (b) All of FMS Department from the of databaseGeophysics, of 194 FacultyDepartment of Science, of Geophysics, University Faculty of Zagreb of (DGFSUZ). Science, University Green rectangle of Zagreb shows (DGFSUZ). the area shownGreen in rectangle part (a). 195 HR:shows Croatia; the area B-H: shown Bosnia in andpart Herzegovina;(a). HR: Croatia; ME: B- Montenegro.H: Bosnia and Herzegovina; ME: Montenegro.

196 2.3. InversionTable 1. ofTime the Macroseismic of occurrence, Field hypocentral coordinates in decimal degrees ( st. errors in km), 197 Macroseismicmagnitudes, and investigations the best double- startedcouple solutions immediately for the after mainshock the mainshock and five aftershocks, occurred. as The obtained collected 198 databy (see various above) agencies permitted and assigningin this study intensity (bold).  to,  145,  are localities. the strike, Macroseismic dip, and rake fieldof the of best the d mainshockouble- 199 was modeledcouple (BDC) using of the theMEEP CMT solutionv.2.0 algorithm or of the [ 37 FMS.] modified 1 DT: Centroid as described time minus in detail hypocentre by [38, 39 time;]. Figure C: 6 200 presentsCentroid Intensity location; Data E: Microseismic Points (IDPs) epicentre in the; BDC epicentral: Best Double area.-Couple Maximum; FMS: intensityThe first P- ofwave VIII polarity MSK was 2 201 observedFocal Mechanism in Ston and Solution. in the villagesUSGS [32], of GCMT Podimoćand [33], ISC [34 Mravinca], ISC-GEM (Figure [35], RCMT1). We[36]. have inverted all

IDPs with intensities Iobs VI MSK for the macroseismic earthquake parameters:1°, 1 Coordinates°, 1° of the Time,≥ Lat., Lon., ° C, Depth, macroseismicNo. Date epicentre (φ , λ ), macroseismic focal depth (hm),Mag. and epicentral2°, 2 intensity°, 2° (SourceI ). The2 DT (s)m1 m °N  E E1 km 0 obtained parameters are: (BDC or FMS)1 Mainshock Mw 6.0, 328, 32, 92 5φ = 42.82520:44:17.3N 3 km, λ = 17.835 E 4 km, h = 6 2 km, I = 8.1 MSK 1 m ◦ 42.78m 17.77 ◦ C 15.0m mb 5.6, 0146, 58, 89 GCMT September, DT = 8.1 s± ± ± MS 6.0 (BDC) In order1996 to homogenise the dataset and reduce observations to the average soil, in the inversion we have reduced the observed intensity in Ston by half a degree of MSK (from285, VIII 55, to 30 VII–VIII MSK), since2 Herak et al. [4]20:44:09 have shown 42.80 that the17.94 damage E to the10.0 building Mwc stock 6.0 in the177, old 66, town141 centreUSGS was

closely related to the estimated soil ampliﬁcation determined by ambient noise(FMS) measurements. The 345, 40, 102 macroseismic epicentre is located close to the Podimo´cvillage, about 10.5 km to the NW from the 3 20:44:11.3 42.78 17.92 E 13.3 Mw 5.9 150, 52, 80 ISC microseismic epicentre where the fault rupture started (Figure6). (FMS)

Remote Sens. 2020, 12, 1157 10 of 23 Remote Sens. 2020, 12, x FOR PEER REVIEW 9 of 22

212 213 Figure 6.Figure Macroseismic 6. Macroseismic map of map the ofepicentral the epicentral area area of the of theSton Ston–Slano–Slano mainshock, mainshock, 55 September,September, 1996. 1996. Intensities according to the Medvedev–Sponheuer–Karnik (MSK) scale. The inset shows the 214 Intensities according to the Medvedev–Sponheuer–Karnik (MSK) scale. The inset shows the geographical position of the area shown. 215 geographical position of the area shown. 3. DInSAR Observations 216 In orderThe to homogeni Differential Interferometricse the dataset Synthetic and reduce Aperture observations Radar (DInSAR) to the technique average has soil, been in widely the inversion 217 we have usedreduced for investigation the observed and characterisationintensity in Ston of coseismic by half a surface degree displacements of MSK (from [40]. VIII The techniqueto VII–VIII is MSK), 218 since Herakbased et on al. the [4] measurement have shown of thethat di fftheerential damage phase to angle the between building two stock coherent in the Synthetic old town Aperture centre was 219 Radar (SAR) acquisitions over the same area. After subtraction of the topographic phase component closely relatedfrom the to obtained the estimated phase diff erence,soil amplification the technique candetermined determine the by ground ambient surface noise motion measurements. caused by The 220 macroseismicthe earthquake epicentre in theis radarlocated Line-Of-Sight close to the (LOS) Podimoć direction. village, about 10.5 km to the NW from the 221 microseismicIn epicentre this study, w wehere used the the fault SAR imagesrupture acquired started by (Figure the C-band 6). (5.66 cm wavelength) European Remote Sensing satellite (ERS-2) before and after the earthquake from both ascending and descending 222 3. DInSARorbits. Observations The coseismic interferograms were formed from the image pairs described in Table2 by using InSAR Scientific Computing Environment (ISCE) software developed by the Jet Propulsion Laboratory 223 The(JPL; Differential [41]). Coseismic Interferometric interferograms Synthetic were generated Aperture from Radar the ERS-2(DInSAR) Single-Look technique Complex has (SLC)been widely 224 used for productsinvestigation obtained and from characterisation the European Space of Agency coseismic (ESA). surface We selected displacements the interferometric [40]. pairsThe techniqu with e is 225 based ona the favourable measurement perpendicular of the baseline differen (Bperptial) phase smaller angle than 60 between m to suppress two coherent the topographic Synthetic phase Aperture contribution in the interferometric phase. The height of ambiguity of the ascending and descending 226 Radar (SAR) acquisitions over the same area. After subtraction of the topographic phase component 227 from the obtained phase difference, the technique can determine the ground surface motion caused 228 by the earthquake in the radar Line-Of-Sight (LOS) direction. 229 In this study, we used the SAR images acquired by the C-band (5.66 cm wavelength) European 230 Remote Sensing satellite (ERS-2) before and after the earthquake from both ascending and 231 descending orbits. The coseismic interferograms were formed from the image pairs described in 232 Table 2 by using InSAR Scientific Computing Environment (ISCE) software developed by the Jet 233 Propulsion Laboratory (JPL; [41]). Coseismic interferograms were generated from the ERS-2 Single- 234 Look Complex (SLC) products obtained from the European Space Agency (ESA). We selected the 235 interferometric pairs with a favourable perpendicular baseline (Bperp) smaller than 60 m to suppress 236 the topographic phase contribution in the interferometric phase. The height of ambiguity of the 237 ascending and descending ERS-2 interferometric pairs gives a 1.8  10–2 cm and 5.1  10–3 cm error in

Remote Sens. 2020, 12, 1157 11 of 23

ERS-2 interferometric pairs gives a 1.8 10–2 cm and 5.1 10–3 cm error in ground displacement × × determination with a 1 m change in topography, respectively. This implies a low sensitivity of the interferometric phase to topographic errors, typically on the order of ~10 m for the Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM) [42]. The Ston–Slano area represents a perfect location for DInSAR application due to infrequent rainfall, as well as sparse vegetation and human cultivation of land, meaning that the long-wavelength tectonic signal can be observed even after a long time. However, the signiﬁcant atmospheric signal contamination of the interferometric phase (erroneous phase change due to perturbations in pressure, temperature, and relative humidity in the lower part of the troposphere, < 5 km) can be still found especially near the coast. In the selection process of the best interferometric pairs, we paid special attention to the potential atmospheric phase contamination by using the pairwise logic [43]. For example, we discarded the interferometric pair between 9 August, 1996 and 27 July, 1997 from descending orbit with a Bperp of –41 m from this study, as we found that the phase of SAR image acquired on the 27 July, 1997 was strongly aﬀected by atmospheric noise.

Table 2. Interferometric pairs used in diﬀerential interferometric synthetic aperture radar (DInSAR) processing (mainshock occurred on 5 September, 1996). Bperp and Btemp represent perpendicular and temporal baseline of DInSAR pairs respectively, whereas Hamb is height of ambiguity.

Orbit Master Image Slave Image Sat. Track B (m) B (days) H (m) Direction (yr/mm/dd) (yr/mm/dd) perp temp amb ERS-2 Descending 451 09/08/1996 16/051997 –59 280 158 ERS-2 Ascending 501 06/11/1995 01/091997 –17 665 547

The DInSAR processing can be divided into several steps: coregistration, interferogram generation, flat-phase, and topographic phase contribution corrections, adaptive filtering, phase unwrapping, and geocoding. More details about DInSAR processing steps can be found in Hanssen [44]. We applied the precise orbital information provided by the Delft Institute for Earth-Oriented Space Research in the Netherlands [45] for the coregistration of the SAR SLC images and flat-phase contribution removal from the interferometric phase. Tropospheric phase correction was applied by removing the tropospheric phase contribution simulated on the 1 arc s (30 m) SRTM DEM. After tropospheric phase correction, an adaptive Goldstein filter [46] with alpha value of 0.7 was used to further reduce the interferogram phase noise. The obtained filtered interferometric phase wrapped by the 2π moduli was unwrapped with a minimum cost-flow SNAPHU algorithm [47] to obtain relative LOS surface displacements and then geocoded to a WGS84 geographic coordinate system. Due to relatively high coherence (0.65 ± 0.24; Figure 8b), the obtained interferograms are considered to be reliable. We applied a multilook ratio of 1:5 for range and azimuth directions, respectively, to obtain ~20 m pixel posting of the geocoded unwrapped interferograms. The final results are the geocoded coseismic surface displacement fields in the LOS directions of ERS-2 ascending and descending orbits (see Figures7 and8). The displacement fields represent LOS surface movements with respect to the reference point (unwrapping seed point), located arbitrarily outside of the deformation zone. Figure7a,c shows a very spatially complex coseismic displacement field with several sets of apparently overlapping concentric fringe patterns, which indicates that multiple faults and/or fault segments may have ruptured during the earthquake sequence. Both interferograms include the postseismic period comprising two Mw > 5.0 aftershocks (Table1 and Figure5a) that could have interfered with the mainshock’s displacement field. Remote Sens. 2020, 12, 1157 12 of 23 Remote Sens. 2020, 12, x FOR PEER REVIEW 11 of 22

273 274 Figure 7. (a) European radar satellite (ERS-2) ascending orbit wrapped interferogram of the coseismic Figure 7. (a) European radar satellite (ERS-2) ascending orbit wrapped interferogram of the 275 displacement induced by the Ston–Slano earthquake. (b) ERS-2 ascending orbit unwrapped coseismic displacement induced by the Ston–Slano earthquake. (b) ERS-2 ascending orbit unwrapped 276 interferogram of the coseismic displacement. (c) ERS-2 descending orbit wrapped interferogram of interferogram of the coseismic displacement. (c) ERS-2 descending orbit wrapped interferogram of 277 the coseismic displacement. (d) ERS-2 descending orbit unwrapped interferogram of the coseismic the coseismic displacement. (d) ERS-2 descending orbit unwrapped interferogram of the coseismic 278 displacement. Line-of-sight (LOS) surface displacements in (a) and (c) are wrapped by the modulus displacement. Line-of-sight (LOS) surface displacements in (a) and (c) are wrapped by the modulus 2π, 279 2π,where where one one colour colour cycle cycle represents represents ground ground motion motion of 2.8of 2.8 cm cm (the (the satellite’s satellite’s beam beam half half wavelength). wavelength). Black 280 Blacksquares squares in the in upper the upper left corners left corners mark themark reference the reference point for point the phasefor the unwrapping phase unwrapping located outsidelocated of 281 outsidethe deformation of the deform zone.ation zone.

282 FigureThe fringes 7a and patterns 7c show shows a very the groundspatially movement complex coseismic towards thedisplacement satellite in field the LOS with directions several sets that 283 ofspans apparently over the overlapping area about concentric 19.2 km by fringe 11.6 km patterns, extending which NW indicates from Slano. that multiple The main faults diﬀerence and/or between fault 284 segmentsthe ascending may have(Figure ruptured7a) and the during descending the earthquake (Figure7c) sequence. orbit interferograms Both interferograms is a clear absence include of the the 285 postseismicfringe (pattern period P1a, co Figuremprising8a) intwo the M descendingw > 5.0 aftershocks one. This (Table is consistent 1 and Figure with the 5a) suggestion that could that have the 286 interferedcoseismic with ground the displacementmainshock’s displacement was generated field. by a transpressional left-lateral fault with a large reverse 287 componentThe fringe (e.g., patterns FMS No.show 2 andthe ground 6 in Table movement1 and Figure towards5a) dipping the satellite to the in NE, the in LOS which directions the hanging that 288 spanswall over of the the causative area about fault 19.2 moves km by up 11.6 and km to extending the NW, towardsNW from the Slano. satellite The main in its difference ascending between trajectory 289 the(Figure ascending7a). The (Figure maximum 7a) and observed the descending LOS displacement (Figure 7c) of orbit ~38 interferograms cm (Figure7b) and is a ~30clear cm absence (Figure of7d) the is 290 fringefound (pattern near the P1a Podimo´cvillage, Figure 8a) in (patternthe descending P5, Figure one.8a) This in the is ascendingconsistent andwith descending the suggestion interferogram, that the 291 coseismicrespectively. ground Its location displacement lies about was 11 generated km to the by NW a transpressional of the microseismic left-lateral epicentre fault determined with a large here 292 reverse(No. 6 component in Table1 and (e.g. Figure, FMS5a), No. but 2 coincidesand 6 in Table with the1 and inverted Figure position 5a) dipping of the to macroseismic the NE, in which epicentre the 293 hanging(Figures wall5 and of7 ),the i.e., ca theusative location fault of moves maximum up and damage. to the NW, Although towards half the of patternsatellite P5 in is its decorrelated ascending 294 trajectory(Figure8 b),(Figure the other 7a). The half maximum shows a high observed fringe gradientLOS displacement (Figure8a) of that 38 suggests cm (Figure that 7b) most and of  the30 slipcm 295 (Figureoccurred 7d)on is found a shallow near fault the Podimoć segment extendingvillage (pattern to the surfaceP5, Figure with 8a a) in possible the ascending surface ruptureand descending associated 296 interferogram,with the observed respectively. decorrelation. Its location lies about 11 km to the NW of the microseismic epicentre 297 determined here (No. 6 in Table 1 and Figure 5a), but coincides with the inverted position of the 298 macroseismic epicentre (Figures 5 and 7), i.e., the location of maximum damage. Although half of 299 pattern P5 is decorrelated (Figure 8b), the other half shows a high fringe gradient (Figure 8a) that 300 suggests that most of the slip occurred on a shallow fault segment extending to the surface with a 301 possible surface rupture associated with the observed decorrelation.

Remote Sens. 2020, 12, 1157 13 of 23 Remote Sens. 2020, 12, x FOR PEER REVIEW 12 of 22

302

303 FigureFigure 8. (a )8. Black (a) Black lines lines (D1–D4) (D1–D4) lie lie along along the the fringefringe discontinuities observed observed in both in both interferograms interferograms 304 (Figure(Figure7a,c; only7a,c; only the ascendingthe ascending interferogram interferogram is is shown shown here). Long Long,, dashed dashed line line indicates indicates assumed assumed 305 continuationcontinuation of the of discontinuitythe discontinuity off offshore.shore. Short Short dashes emphasise emphasise less less clearly clearly expressed, expressed, secondary secondary 306 discontinuities.discontinuities. None None of theof the observed observed discontinuities discontinuities correlates correlates with with topography. topography. White White tags P1 tags–P5 P1–P5 307 specifyspecify the fringe the fringe patterns. patterns. The The blue blue and and green green crosses mark mark the the macroseismic macroseismic and and microseismic microseismic 308 epicentreepicentre of the of mainshock,the mainshock, respectively. respectively. ‘A’ ‘A’ marks marks the pattern pattern deformation deformation close close to the to microseismic the microseismic 309 epicentre. (b) Coherence of ERS-2 ascending orbit unwrapped interferograms with value range (0.65 epicentre. (b) Coherence of ERS-2 ascending orbit unwrapped interferograms with value range 310 ± 0.24). (0.65 0.24). ± 311 4. Geological Observations 4. Geological Observations 312 The structural-geological analysis of the Ston–Slano area in terms of fault kinematics 313 Thedetermination, structural-geological i.e., relation analysis between of observable the Ston–Slano fault structures area in terms in respect of fault tokinematics the past and determination, present 314i.e., relationstress fields, between was mainly observable focused fault along structures the mapped in respect faults at to the the contacts past and and present within the stress Mesozoic fields, was mainly focused along the mapped faults at the contacts and within the Mesozoic and Eocene successions, areas of relocated microseismic and macroseismic epicentres, as well as some specific DInSAR observations (indicating previously unknown surface discontinuities, e.g., discontinuity D4; Figures2, 8 and9). In the studied area (about 25 km long and 10 km wide; Figure9) structural data were collected Remote Sens. 2020, 12, x FOR PEER REVIEW 13 of 22

315 andRemote Eocene Sens. 2020 successions,, 12, 1157 areas of relocated microseismic and macroseismic epicentres, as well as14 some of 23 316 specific DInSAR observations (indicating previously unknown surface discontinuities, e.g., 317 discontinuityon outcrop-scale D4; Figures fault planes 3, 8, and (e.g., 9). dip In directionthe studied/angle area of (about fault planes25 km andlong orientation and 10 km ofwide; carbonate Figure 318 9)slickensides structural data including were theircollected azimuth on outcrop/plunge-scale and sense fault ofplanes movement). (e.g., dip Collected direction/angle structural of datafault forplanes 105 319 andfault orientation planes on of 129 carbonate locations slickensides (Figure9) were including separated their into azimuth/plunge kinematically compatibleand sense of datasets movement). and 320 Collectedprocessed structural by Tectonics data FP softwarefor 105 fault[48]. Using planes the on P–T 129 axis locations method (Figure [49,50] the 9) theoretical were separated maximum into 321 kinematically(σ1), intermediate compatible (σ2), and dat minimumasets and stress processed axes ( σby3) Tectonics were calculated, FP software whereas [48]. palaeo-synthetic Using the P–T focal axis 322 methodmechanisms [49,50] for the analysed theoretical fault maximum segments were(σ1), determinedintermediate as (σ2) representatives, and minimum of the stress palaeostress axes (σ3) ﬁelds were 323 calculated,with the Right whereas Dihedra palaeo Method-synthetic [51]. focal mechanisms for analysed fault segments were determined 324 as representatives of the palaeostress fields with the Right Dihedra Method [51].

325

326 FigureFigure 9. 9. SimplifiedSimpliﬁed structural structural map map of of the the Ston Ston–Slano–Slano area area with with faults faults mapped mapped onshore and offshore. oﬀshore. 327 FaultFault tr tracesaces are are generali generalisedsed after after [10,11, [10,11,19].19]. Fault Fault measureme measurementsnts conducted at the 127127 locationslocations areare 328 indicatedindicated with with multicolour multicolour dots dots in in accordance accordance to to the the measured measured fault’s kinematic properties.properties. LocationsLocations 329 ofof microseismic microseismic and and macroseismic macroseismic epicentres epicentres of of the 1996 Ston–SlanoSton–Slano earthquake are are indicated indicated in in 330 accordanceaccordance to to Figure Figure 23,, whereas purple lineslines areare thethe fringefringe discontinuitiesdiscontinuities interpretedinterpreted inin FigureFigure8 .8. 331 Abbreviations:Abbreviations: PF PF:: Pelješac Pelješac fault; fault; DF DF:: Doli Doli fault; fault; SF SF:: Slano fault; MF:MF: Mravinca fault.fault.

Structural measurements of the mapped fault planes within the Ston–Slano area indicated both 332 Structural measurements of the mapped fault planes within the Ston–Slano area indicated both dip-slip and strike-slip fault kinematics, whereas observed oblique-slip kinematics were characterised 333 dip-slip and strike-slip fault kinematics, whereas observed oblique-slip kinematics were by a dominant reverse dip-slip component. Positioned in heavily tectonised zones, mapped fault 334 characterised by a dominant reverse dip-slip component. Positioned in heavily tectonised zones, planes were separated into four compatible structural datasets (Figure3). Accordingly, based on the 335 mapped fault planes were separated into four compatible structural datasets (Figure 2). Accordingly, kinematic criteria, 83 measured fault planes were characterised by dip-slip kinematics, out of which 336 based on the kinematic criteria, 83 measured fault planes were characterised by dip-slip kinematics, 59 have reverse and 24 normal faulting kinematics. Twenty-two fault planes accommodated either 337 out of which 59 have reverse and 24 normal faulting kinematics. Twenty-two fault planes oblique-slip or strike-slip motion. The aforementioned fault kinematic groups were subdivided into 338 accommodated either oblique-slip or strike-slip motion. The aforementioned fault kinematic groups compatible fault groups and fault group subsets (Figure 10 and Table3). 339 were subdivided into compatible fault groups and fault group subsets (Figure 10 and Table 3).

Remote Sens. 2020, 12, x FOR PEER REVIEW 14 of 22

340 Results of structural analysis show that reverse fault planes can be separated into two fault 341 groups and four subsets. The first reverse fault group (RF-1; Figure 10a) is characterised by conjugate 342 fault pairs, i.e., fault subsets characterised by the NW‒SE strike, dipping both towards the NE and 343 SW (dip angles are 51 and 32°, respectively; Table 3), indicating NE- and SW-directed tectonic 344 transport (Figures 10a and 11). Structural analysis of the palaeostress field using the P–T axis method 345 [50,51] and derived synthetic structural focal mechanisms (Figure 10a), indicated that the observed 346 palaeostress compressional field is associated with a P-axis predominantly trending NE‒SW (see 347 Table 3 for details). The second group of the observed reverse faults (RF-2; Figure 10b) is generally 348 characterised by conjugate fault pairs dipping towards ESE and WNW and WNW- and ESE-directed 349 tectonic transport, respectively. Computed representative palaeostress field indicates a 350 compressional palaeostress field associated with a P-axis dominantly trending NW‒SE (Figure 10b 351 and Table 3). In addition to typical compressional structures, i.e., reverse fault planes, 22 oblique-slip 352 or strike-slip fault planes were measured in this study and were grouped into the oblique-slip fault 353 group (OF; Figure 10c). With dominant N–S (locally NW–SE and NE–SW) and E–W strike (see Figure 354 10c and Table 3) and subvertical to subhorizontal geometry, respectively, observed subvertical and 355 sinistral faults are kinematically linked with prevalence of transpressional tectonic phase (Figure 10c) Remote Sens. 2020, 12, 1157 15 of 23 356 characterised by NW–SE trending P-axis that gently dips (30°) towards NW (see Figure 10c and Table 357 3 for details).

Remote Sens. 2020, 12, x FOR PEER REVIEW 15 of 22

358

359 FigureFigure 10. StructuralStructural diagrams diagrams for for the Ston Ston–Slano–Slano fault zone. The The white white quadrants quadrants on on the structural 360 372 beachbeach-ball-ball diagrams diagrams represent represent compression, compression, while while the shaded the shaded quadrants quadrants represent represent tension. tension. (a) Reverse (a) 361 373 Reversefault groupFigure fault 1group11. (RF-1); Photograph 1 (RF (b-)1) Reverse(a; )( bof) Reversethe reverse fault fault group fault group plane 2 (RF-2); (Fp2 (RF = -16/55;2) (c;) (c Oblique)indicating Oblique faultSW-directed fault group group tectonic (oblique-slip(oblique -slip or or 374 transport) with detail of visible reverse motion indicators, i.e., fault striations (b) within the Upper 362 strikestrike-slip)(OF);-slip)(OF); ((dd)) NormalNormal fault fault group group (NF). (NF) The. The red dots, red dots, black black rectangles, rectangles, and blue and triangles blue triangles indicate 375 Cretaceous carbonates c. 3 km SW of Slano (Date: 7 March, 2018; Location: 42.76 °N, 17.88 °E; σ σ σ 363 376 indicate1, 2, andPhotographed σ1, σ32,, and respectively. σ by:3, respectively.Bojan Matoš). Table 3. Mean geometric properties of the observed fault planes within the Ston–Slano fault zone with 364 377 In the StonTable– 3.Slano Mean geometricarea, within properties the ofmapped the observed fault fault zones planes 24within normal the Ston–Slano fault planes fault zone were measured 378 calculatedwith kinematiccalculated kinematic indicators indicators and parameters. and parameters. Fault Fault planes planes were delineated delineated with with respect respect to to their 365 (NF; Figure 10d). Observed NW–SE striking fault subsets are dipping both towards NE an SW, at the 379 geometricaltheir geometrical properties properties and kinematic and kinematic compatibility compatibility within within the thefollowing following groups groups (see (see Figure Figure 10): 10): RF-1: 366 dip380 angles Reverse ofRF-1: fault 51 Reverseand group 34°, fault 1; RF-2:respectively group Reverse 1; RF-2: faultReverse(Figure group fault 10 2;grd;oup OF: Table 2; Oblique OF: 3). Oblique At fault the fault group same group (oblique-slip time, (oblique-slip structuralor or strike-slip); analysis of 367 the381 palaeostress and NF:strike-slip); Normal field, and fault i.e. NF:,group. P Normal–T axis Fault fault method group. types: Fault, R:as Reverse;well types: as R: derivedReverse; O-S: Oblique-slip; O-S: synthetic Oblique-slip; S-S: structural S-S: Strike-slip; Strike-slip; focal N: mechanisms Normal. 382 N: Normal. Orientation of the P- and T-axes are based on constructed synthetic structural beach-ball 368 (FigureOrientation 10d and of Table the P- 3) and indicated T-axes are a basedlocally on expressed constructed extensional synthetic structural kinematic beach-ball tectonic diagrams. phase in the 383 diagrams. 369 study area. This tectonic phase was associated to subvertical P-axis (orientation of P-axis is 173/75; 370 see Table 3) which resulted in the general E–W local extension of the observed structures. No. Striation P-axis T-axis Dip Dip 371 Fault Fault of Pitch Fault azimuth angle Trend Plunge Trend Plunge Trend Plunge group subset fault (°) type (°) (°) (°) (°) (°) (°) (°) (°) data

RF-1a 27 44 51 69 113 49 RF-1 R 50 8 205 82 RF-1b 15 216 32 69 237 49

RF-2a 11 95 52 57 139 46 RF-2 R 100 8 247 80 RF-2b 6 291 32 51 218 46

OFa 16 86 88 3 O-S 138 4 OF 296 30 70 50 OFb 6 184 20 0 S-S 100 1

NFa 18 41 51 69 85 51 NF N 173 75 51 08 NFb 6 243 4 55 221 49 384

385 Results5. Discussion of structural analysis show that reverse fault planes can be separated into two fault groups and four subsets. The first reverse fault group (RF-1; Figure 10a) is characterised by conjugate fault 386 As presented above, the observed coseismic displacement field of the Ston–Slano earthquake pairs,387 i.e.,sequence fault is subsets surprisingly characterised complex. It suggests by the that NW-SE at least strike, four, but dipping possibly bothas many towards as nine distinct the NE and SW (dip388 anglesfault aresegments 51 and ruptured 32◦, respectively; during the sequence Table (Figur3), indicatinge 8a). It is unfortunate NE- and SW-directedthat no coherent tectonic DInSAR transport (Figures389 interferograms 10a and 11). generated Structural with analysis a second of SAR the image palaeostress acquired shortly field using after the the mainshock P–T axis could method have [ 50,51] and derived390 been synthetic obtained. structural The only, marginally focal mechanisms useful interfer (Figureogram, 10 is a),made indicated using a SAR that acquisition the observed from the palaeostress 391 descending orbit of 13 September, 1996 (Figure 12a), i.e., eight days after the mainshock. Comparing compressional field is associated with a P-axis predominantly trending NE-SW (see Table3 for details). 392 it to Figure 12b (also an interferogram generated with SAR acquisitions from the descending orbit), 393 it is clear that nearly all main features of the coseismic surface deformation are already present on the

Remote Sens. 2020, 12, 1157 16 of 23

The second group of the observed reverse faults (RF-2; Figure 10b) is generally characterised by conjugate fault pairs dipping towards ESE and WNW and WNW- and ESE-directed tectonic transport, respectively. Computed representative palaeostress ﬁeld indicates a compressional palaeostress ﬁeld associated with a P-axis dominantly trending NW-SE (Figure 10b and Table3). In addition to typical compressional structures, i.e., reverse fault planes, 22 oblique-slip or strike-slip fault planes were measured in this study and were grouped into the oblique-slip fault group (OF; Figure 10c). With dominant N–S (locally NW–SE and NE–SW) and E–W strike (see Figure 10c and Table3) and subvertical to subhorizontal geometry, respectively, observed subvertical and sinistral faults are kinematically linked with prevalence of transpressional tectonic phase (Figure 10c) characterised by NW–SE trending P-axis that gently dips (30◦) towards NW (see Figure 10c and Table3 for details). Remote Sens. 2020, 12, x FOR PEER REVIEW 15 of 22

372 373 Figure 11. Photograph (a) of the reverse fault plane (Fp = 16/55; indicating SW-directed tectonic Figure 11. Photograph (a) of the reverse fault plane (Fp = 16/55; indicating SW-directed tectonic 374 transport) with detail of visible reverse motion indicators, i.e., fault striations (b) within the Upper transport) with detail of visible reverse motion indicators, i.e., fault striations (b) within the Upper 375 Cretaceous carbonates c. 3 km SW of Slano (Date: 7 March, 2018; Location: 42.76 °N, 17.88 °E; Cretaceous carbonates c. 3 km SW of Slano (Date: 7 March, 2018; Location: 42.76 N, 17.88 E; 376 Photographed by: Bojan Matoš). ◦ ◦ Photographed by: Bojan Matoš).

377 Table 3. Mean geometric properties of the observed fault planes within the Ston–Slano fault zone In the Ston–Slano area, within the mapped fault zones 24 normal fault planes were measured 378 with calculated kinematic indicators and parameters. Fault planes were delineated with respect to (NF; Figure 10d). Observed NW–SE striking fault subsets are dipping both towards NE an SW, at the 379 their geometrical properties and kinematic compatibility within the following groups (see Figure 10): dip angles of 51 and 34◦, respectively (Figure 10d; Table3). At the same time, structural analysis of 380 RF-1: Reverse fault group 1; RF-2: Reverse fault group 2; OF: Oblique fault group (oblique-slip or the palaeostress ﬁeld, i.e., P–T axis method, as well as derived synthetic structural focal mechanisms 381 strike-slip); and NF: Normal fault group. Fault types: R: Reverse; O-S: Oblique-slip; S-S: Strike-slip; (Figure 10d and Table3) indicated a locally expressed extensional kinematic tectonic phase in the 382 N: Normal. Orientation of the P- and T-axes are based on constructed synthetic structural beach-ball study area. This tectonic phase was associated to subvertical P-axis (orientation of P-axis is 173/75; see 383 diagrams. Table3) which resulted in the general E–W local extension of the observed structures.

5. Discussion No. Striation P-axis T-axis Dip Dip AsFault presented Fault above,of the observed coseismicPitch Fault displacement ﬁeld of the Ston–Slano earthquake azimuth angle Trend Plunge Trend Plunge Trend Plunge sequencegroup is surprisinglysubset fault complex. It suggests(°) thattype at least four, but possibly as many as nine distinct (°) (°) (°) (°) (°) (°) (°) (°) fault segments ruptureddata during the sequence (Figure8a). It is unfortunate that no coherent DInSAR interferograms generated with a second SAR image acquired shortly after the mainshock could have been obtained.RF The-1a only,27 marginally44 51 useful 69 interferogram, 113 is made49 using a SAR acquisition from the RF-1 R 50 8 205 82 descending orbitRF-1b of 1315 September, 216 199632 (Figure69 12a), i.e., eight237 days49 after the mainshock. Comparing it to Figure 12b (also an interferogram generated with SAR acquisitions from the descending orbit), it is clear that nearlyRF-2a all11 main features95 of52 the coseismic57 surface139 deformation46 are already present on the RF-2 R 100 8 247 80 ERS-2 acquisitionRF-2bon 136 September,291 1996.32 This51 implies that,218 for instance,46 the strongest aftershock (17 September, 1996, 13:45, Mw 5.4, see Table1) had practically no role in formation of the ﬁnal pattern of the coseismicOFa surface 16 deformation 86 obtained88 after3 theO-S earthquake 138 sequence4 ended, i.e., observed in the OF 296 30 70 50 interferogramsOFb in Figure 6 7. In184 the period20 before0 13 September,S-S 100 19961 there was only one event that could have caused ground deformation large enough to be measured with DInSAR—the one of 9 September, NFa 18 41 51 69 85 51 1996 (Mw 5.3, see Table1). The expected fault rupture length and rupture area for a magnitude M w 5.3 NF N 173 75 51 08 NFb 6 243 4 55 221 49 384

385 5. Discussion 386 As presented above, the observed coseismic displacement field of the Ston–Slano earthquake 387 sequence is surprisingly complex. It suggests that at least four, but possibly as many as nine distinct 388 fault segments ruptured during the sequence (Figure 8a). It is unfortunate that no coherent DInSAR 389 interferograms generated with a second SAR image acquired shortly after the mainshock could have 390 been obtained. The only, marginally useful interferogram, is made using a SAR acquisition from the 391 descending orbit of 13 September, 1996 (Figure 12a), i.e., eight days after the mainshock. Comparing 392 it to Figure 12b (also an interferogram generated with SAR acquisitions from the descending orbit), 393 it is clear that nearly all main features of the coseismic surface deformation are already present on the

Remote Sens. 2020, 12, x FOR PEER REVIEW 16 of 22

394 ERS-2 acquisition on 13 September, 1996. This implies that, for instance, the strongest aftershock (17 395 September, 1996, 13:45, Mw 5.4, see Table 1) had practically no role in formation of the final pattern 396 of the coseismic surface deformation obtained after the earthquake sequence ended, i.e., observed in 397 theRemote interferograms Sens. 2020, 12 ,in 1157 Figure 7. In the period before 13 September, 1996 there was only one event that17 of 23 398 could have caused ground deformation large enough to be measured with DInSAR—the one of 9 399 September, 1996 (Mw 5.3, see Table 1). The expected fault rupture length and rupture area for a 2 earthquake are 3–5 km and 15–20 km , respectively (according2 to [52]), so it is possible that at least 400 magnitude Mw 5.3 earthquake are 3–5 km and 15–20 km , respectively (according to [52]), so it is 401 possibleone of that the observed at least one fringe of discontinuities the observed fringe shown discontinuities in Figure8a is due shown to the in aftershockFigure 8a is of due 9 September, to the 402 aftershock1996. Given of 9 September the position, 1996 of its. Given epicentre the position (Table1 andof its Figure epicentre5) and (Table taking 1 and into Figure account 5) and rather taking large 403 intoconﬁdence account rather intervals large for confidence its best location, intervals the for most its best likely location, candidate the most is pattern likely P4 candidate that is related is pattern to the 404 P4 Dolithat faultis related (Figures to 8thea and Doli9). fault However, (Figures as even8a and stronger 9). However, aftershock as ofeven 17 September,stronger aftershock 1996 apparently of 17 405 Septemberdid not cause, 1996 ground apparently displacements did not cause observable ground with displacements DInSAR, and observableknowing that with recent DInSAR events, andof this 406 knowingsize in the that greater recent area events were of not this recognized size in the in thegreater DInSAR area interferograms, were not recognized this will in have the to DInSAR remain an 407 interferograms,unsolved issue. this will have to remain an unsolved issue.

408 409 FigureFigure 12. 12.ERSERS-2-2 descending descending orbit orbit wrapped wrapped interferograms interferograms of the of the coseismic coseismic displacement displacement obtained obtained 410 from interferometric pair; (a) 25 August, 1995–13 September, 1996 (Bperp 108 m); (b) 9 August, 1996– from interferometric pair; (a) 25 August, 1995–13 September, 1996 (Bperp 108 m); (b) 9 August, 1996–16 411 16 May, 1997 (Bperp -59 m). LOS surface displacements are wrapped by the modulus 2π, where one May, 1997 (Bperp -59 m). LOS surface displacements are wrapped by the modulus 2π, where one colour 412 colourcycle cycle represents represents ground ground motion motion in a sizein a size of the of satellite’sthe satellite’s beam beam half wavelengthhalf wavelength (2.8 cm).(2.8 cm).

413 TakingTaking all allstated stated above above into into consideration, consideration, we we are are left left with with a case a case of ofapparently apparently very very complex complex 414 multiplemultiple fault fault rupture rupture during during the the main main event event of 55 September,September, 1996. 1996. It It started started about about 4.5 4.5 km km ESE ESE of Slanoof 415 Slanonear near the the Majkovi Majkovi village village (Figure (Figure6), at 6), the at depththe depth of about of abou 11 kmt 11 (Tablekm (Table1), and 1), proceededand proceeded NW-wards, NW- 416 wards,either either along along the reverse the reverse Slano Slano fault (SF)fault or (SF) the or Pelješac the Pelješac fault (PF) fault (Figures (PF) (Figures8a,9 and 8a 13, 9), whoseand 13) traces whose are 417 tracesmapped are mapped about 4 andabout 9 km 4 and to the 9 km SSW, to respectively. the SSW, respectively. Both of their Both strikes of their agree strikes well with agree the well strike with of one 418 theof strike the nodal of one planes of the of nodal the FMS planes (φ1 of= the293 ◦FMS, see ( Table1 = 293°,1). Assuming see Table the 1). dipAssuming of the fault the dip as δof1 =the53 fault◦ and a 419 as focal1 = 53° depth and ofa focal 11.4 depth5 km of (Table 11.41 ), 5 and km also(Table taking 1), and the also uncertainty taking the of theuncertainty epicentral of location the epicentral ( 5 km) ± ± 420 locationof the ( FMS 5 km) parameters of the FMS into account,parameters neither into ofaccount, the faults neithe canr be ofgiven the faults a definite can be preference. given a definite However, 421 preference.inspection However, of Figure inspection8a hints that of Figure the fringes 8a hints of thethat P1a the andfringes P2 patternsof the P1a stop and at P2 the patterns discontinuities stop at 422 theD1 discontinuities and D2, which D1 correspondsand D2, which to thecorrespond Slano faults to asthe mapped Slano fault inthe as mapped field (Figures in the8 afield and (Figure9), so wes 423 8a suggestand 9), so that we it suggest is probably that it the is Slanoprobably fault the where Slano the fault main where rupture the main started. rupture The started. near-field The coseismic near- 424 fieldground coseismic deformation ground deformation in the immediate in the vicinity immediate of the vicinity epicentre of the is clearly epicentre seen is in clearly the ascending seen in the orbit 425 ascendinginterferogram orbit interferogram(Figure8a, mark (Figure ‘A’). Left-lateral 8a, mark transpressive ‘A’). Left-lateral rupture transpressive proceeded rupture NW-wards proceeded for about 426 NW8- kmwards (dashed for about part of8 km D1 in(dashed Figure part8a), producingof D1 in Figure the main 8a), lobesproducing P1a (Figure the main8a). lobes Their P1a wide (Figure spatial 427 8a).extent Their in wide the hangingspatial extent wall suggestsin the hanging that the wall rupture suggests was that confined the rupture to the deepwas confined parts ofthe to the fault. deep 428 parts of the fault. 429 After that, the rupture enters the area of the largest observed damage, characterised also by the 430 largest cumulative ground uplift (Figure 7b,d). The very dense fringe pattern P5, partly decorrelated 431 (Figure 8a,b) suggests that most accumulated stress was released closer to the surface (in agreement

Remote Sens. 2020, 12, x FOR PEER REVIEW 17 of 22

432 with the macroseismic focal depth of 6 km), which (according to [53]) can result with potential surface 433 rupture and secondary effects, e.g., liquefaction, building damage, etc. As the earthquake occurred 434 24 years ago, it is not possible to confirm the cause of the decorrelation in the macroseismic epicentral 435 area. Interferograms in this area strongly indicate that several imbricated reverse faults (e.g., the 436 Mravinca Fault; see Figures 9 and 13) were activated by the main rupture. The most clear case is the 437 one of the fringe discontinuity D3a associated with pattern P3 (Figures 7 and 8a; maximum LOS 438 displacements of 19 and 14 cm in ascending and descending interferograms, respectively) and 439 identified in the field as a NW-striking reverse Mravinca fault, an approximately 10 km long segment 440 of the regionally significant High Karst Nappe (MF in Figure 9). Similarly, the smaller or less clearly 441 expressed ruptures (short dashes in Figure 8a) that cannot be attributed to any known aftershock (or 442 indeed to any of the mapped faults), were also probably activated by the main event. Some of them 443 strike at large angles or even perpendicularly to the main mapped faults, indicating complex stress 444 field,Remote and Sens. activation2020, 12, 1157 of a set of smaller linked faults, i.e., conjugate faults within the system. A similar18 of 23 445 observation related to the Ridgecrest earthquake was published by Ross et al. [54].

446 447 FigureFigure 13. 13. ThreeThree profiles proﬁles A A–B,–B, C C–D,–D, E E–F–F (black (black lines) lines) over over ( (aa)) ascending ascending and and ( (bb)) descending descending track track 448 coseismiccoseismic interferograms, interferograms, with with relocated relocated earthquakes earthquakes and and documented documented faults faults depicted depicted with grey withcircles grey and black/purple lines, respectively. Only earthquake locations characterised by a standard error

of the epicentre σ 5 km and with azimuthal coverage characterised by a gap γ 150 are used. H ≤ ≤ ◦ Subplots (c), (d), and (e) show DInSAR coseismic displacements, earthquake hypocentres, and points of cross-sections with the documented faults (PF, DF, SF, MF) 5 km around the profiles. Red and blue circles show the mean coseismic displacement from ERS-2 ascending and descending track interferograms along the profiles, respectively. Relocated earthquakes are shown by different colours according to their magnitude, and topography elevation is shown in light grey. Dark grey patch approximately shows the cross-section of the domain of imbricated reverse faults dipping to the NE. The FMS of the mainshock in (a) and (b) is shown as horizontal equal-area projection on the lower hemisphere. In (e) the view of the focal sphere is horizontally from SE, perpendicularly to the strike of the transect E–F.

After that, the rupture enters the area of the largest observed damage, characterised also by the largest cumulative ground uplift (Figure7b,d). The very dense fringe pattern P5, partly decorrelated Remote Sens. 2020, 12, 1157 19 of 23

(Figure8a,b) suggests that most accumulated stress was released closer to the surface (in agreement with the macroseismic focal depth of 6 km), which (according to [53]) can result with potential surface rupture and secondary effects, e.g., liquefaction, building damage, etc. As the earthquake occurred 24 years ago, it is not possible to confirm the cause of the decorrelation in the macroseismic epicentral area. Interferograms in this area strongly indicate that several imbricated reverse faults (e.g., the Mravinca Fault; see Figures9 and 13) were activated by the main rupture. The most clear case is the one of the fringe discontinuity D3a associated with pattern P3 (Figures7 and8a; maximum LOS displacements of 19 and 14 cm in ascending and descending interferograms, respectively) and identified in the field as a NW-striking reverse Mravinca fault, an approximately 10 km long segment of the regionally significant High Karst Nappe (MF in Figure9). Similarly, the smaller or less clearly expressed ruptures (short dashes in Figure8a) that cannot be attributed to any known aftershock (or indeed to any of the mapped faults), were also probably activated by the main event. Some of them strike at large angles or even perpendicularly to the main mapped faults, indicating complex stress field, and activation of a set of smaller linked faults, i.e., conjugate faults within the system. A similar observation related to the Ridgecrest earthquake was published by Ross et al. [54]. Most probably the main rupture continued further for several kilometres along the fault towards Ston (segment D2 in Figure8a, and the associated pattern P2 with the maximum displacement of about 10 cm), where the final stress release occurred. The rupture of this segment of the Slano fault combined with the soil amplification (see above) could explain the severe damage in Ston, which is 21 km away from the microseismic epicentre. Almost all of the observed DInSAR coseismic displacements are positive (upwards) and confined to the hanging walls of imbricated reverse faults (Figures9 and 13), suggesting coseismic relaxation of deformation accumulated in the faults’ hanging walls by tectonic forces related to the compressional/transpressional stress field, characterised by dominant NE–SW (locally NW–SE) oriented shortening. Virtually complete lack of aftershocks in the most shaken area (outlined in Figure 14), as well as very weak seismicity since 1996 (see Figure4b) imply that the stress release by the mainshock was nearly total. The expected subsurface rupture length (L) for an Mw 6.0 earthquake is between 11.5 and 14.0 km (according to [52]), which roughly matches the length of the D1 line, as shown in Figure8a. Assuming that the centroid location corresponds to the area of the largest observed displacement (Figure7b,d), and considering the centroid time is 8.1 s later than the hypocentral time (DT for event No. 1 in Table1), an average rupture velocity of about 1.5 km/s is implied, which is lower than the global average for such events (e.g., [55]). As shown above, there is a strong possibility that several faults were activated during the main rupture, which considerably increases the total rupture length. The scalar seismic moment Mo = Aµu (where A—rupture area, µ = 3.2 1010 Pa—shear modulus of the crust, u—average slip on the fault) of × 1.2 1018 Nm as given in the Global CMT Catalogue [30,31,33], implies the product A u = L W u × × × × = 3.8 107 m3 (where L, W—average rupture length, and width, respectively). Tentatively assuming L × = 25 km along all activated faults and subfaults, and reasonably assuming the mean slip u in the range of 15–25 cm led to estimates of W between 6 and 10 km. As the earthquake started at the depth of about 11 km, and the fringe patterns P1a and P1b (Figure8a) extend well away from the fault, it seems reasonable to assume that most of the slip in the initial part of the rupture was confined to the deeper parts of the fault. Saturated and concentrated fringes in patterns P3 and especially P5 suggest that in the later stages the rupture moved towards the surface, where it also activated a number of smaller faults (Figures8a and9). Remote Sens. 2020, 12, x FOR PEER REVIEW 18 of 22

449 circles and black/purple lines, respectively. Only earthquake locations characterised by a standard 450 error of the epicentre H ≤ 5 km and with azimuthal coverage characterised by a gap  ≤ 150° are used. 451 Subplots (c), (d), and (e) show DInSAR coseismic displacements, earthquake hypocentres, and points 452 of cross-sections with the documented faults (PF, DF, SF, MF) 5 km around the profiles. Red and blue 453 circles show the mean coseismic displacement from ERS-2 ascending and descending track 454 interferograms along the profiles, respectively. Relocated earthquakes are shown by different colours 455 according to their magnitude, and topography elevation is shown in light grey. Dark grey patch 456 approximately shows the cross-section of the domain of imbricated reverse faults dipping to the NE. 457 The FMS of the mainshock in (a) and (b) is shown as horizontal equal-area projection on the lower 458 hemisphere. In (e) the view of the focal sphere is horizontally from SE, perpendicularly to the strike 459 of the transect E–F.

460 Most probably the main rupture continued further for several kilometres along the fault towards 461 Ston (segment D2 in Figure 8a, and the associated pattern P2 with the maximum displacement of 462 about 10 cm), where the final stress release occurred. The rupture of this segment of the Slano fault 463 combined with the soil amplification (see above) could explain the severe damage in Ston, which is 464 21 km away from the microseismic epicentre. 465 Almost all of the observed DInSAR coseismic displacements are positive (upwards) and 466 confined to the hanging walls of imbricated reverse faults (Figures 9 and 13), suggesting coseismic 467 relaxation of deformation accumulated in the faults’ hanging walls by tectonic forces related to the 468 compressional/transpressional stress field, characterised by dominant NE–SW (locally NW–SE) 469 oriented shortening. Virtually complete lack of aftershocks in the most shaken area (outlined in Remote Sens. 2020, 12, 1157 20 of 23 470 Figure 14), as well as very weak seismicity since 1996 (see Figure 4b) imply that the stress release by 471 the mainshock was nearly total.

472

473 Figure 14. EpicentresEpicentres of of reliably reliably located earthquakes of of the StonSton–Slano–Slano sequence overlain on the 474 ascending orbit i interferogramnterferogram from Figure7 7a.a. OnlyOnly locationslocations characterisedcharacterised byby aa standardstandard errorerror ofof thethe epicentre σH 5 km and with azimuthal coverage characterised by a gap γ 150◦ are shown. Thick 475 epicentre H ≤ 5 km and with azimuthal coverage characterised by a gap  ≤ blue dashed line encloses the most shaken area which is virtually aftershock-free. 476 blue dashed line encloses the most shaken area which is virtually aftershock-free. Locations of foci (Figures4 and 13), as well as the DInSAR interferograms suggest that the 477 The expected subsurface rupture length (L) for an Mw 6.0 earthquake is between 11.5 and 14.0 accumulated stress during the Ston–Slano sequence was released by a complex rupture of the system of 478 km (according to [52]), which roughly matches the length of the D1 line, as shown in Figure 8a. imbricated faults (at least the faults PF, DF, SF, and MF; see Figure9) in the length of more than 20 km. As outlined above, the large part of the overall rupture probably occurred during the mainshock itself. Such a complicated coseismic faulting pattern has never before been documented in the Dinarides. However, the Ston–Slano earthquake is the only event in the Dinarides large enough to be analysed by the DInSAR technique, which suggests that multiple faulting and complex surface deformation may be common in the area, but may not be always detectable or measurable. If so, this can present a serious problem in deﬁning realistic hazard scenarios, especially in deterministic hazard assessment.

6. Conclusions This multidisciplinary study revealed a complex pattern of faulting and rupture propagation during the mainshock of the Ston–Slano earthquake sequence (5 September, 1996, Mw = 6.0). In spite of rather scarce seismological data of that time, we were able to identify the faults that ruptured in the first week of the sequence, most probably during the main rupture itself. The fault-plane solution and the microseismic hypocentre suggest that the reverse rupture with a strong left-lateral component probably started on the Slano fault at the depth of about 11 km. Proceeding unilaterally to the NW with the velocity of about 1.5 km/s, the rupture propagated for about 11 km and entered the area of the maximum stress release surrounding the villages of Mravinca and Podimoć,where the highest intensities were observed. DInSAR interferograms suggest that several hanging wall faults were activated here, the largest of which is the Mravinca reverse fault as a segment of the regional High Karst Nappe. Some of those smaller dislocations strike at large angles to the prevailing strike (SE–NW) of major structures. Continuing further towards Ston, the final stress was released as far as 20 km away from the epicentre, about three km to the NE of Ston. The DInSAR interferograms suggest that Remote Sens. 2020, 12, 1157 21 of 23 coseismic ground deformation is observed in the area of about 200 km2, with the maximum LOS ground displacement reaching 38 cm. This area of the southern External Dinarides is characterised by NE–SW trending regional shortening that is partitioned along the closely spaced imbricated reverse fault system. In this imbricated fault system, large dislocation along the most stressed fault surface is likely to induce stress release also within the already stressed faults, mostly in the hanging wall of the primary rupture. If multiple faulting processes of strong events as complex as the one described above are a rule in the area, rather than exception, this could present a serious problem in defining realistic hazard scenarios, especially in deterministic hazard assessment.

Author Contributions: Conceptualization of research, M.G., M.H., B.M. and I.V.; data curation, M.G., M.H., B.M. and I.V.; DInSAR investigations, M.G.; seismological investigations, M.H.; geological investigations, B.M. and I.V.; software, M.G., M.H. and B.M.; visualization, M.G., M.H., B.M. and I.V.; writing—original draft, M.G., M.H., B.M. and I.V.; writing—review and editing, M.G., M.H., B.M. and I.V.; funding acquisition, B.P. and M.H.; project administration. B.P. and M.H. All authors have read and agreed to the published version of the manuscript. Funding: This research was funded by the Croatian Science Foundation under grants IP-01-2018-8944 and IP-2014-09-9666. Acknowledgments: We acknowledge the free access to satellite images provided by the European Space Agency, which enabled us to perform DInSAR investigation of the Ston–Slano earthquake sequence. We also thank the three anonymous reviewers for their very useful comments, which helped us improve the manuscript. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

1. Markušić,S.; Herak, D.; Ivanˇcić,I.; Sović, I.; Herak, M.; Prelogović,E. Seismicity of Croatia in the period 1993–1996 and the Ston–Slano earthquake of 1996. Geofizika 1998, 15, 83–101. 2. Herak, M.; Herak, D.; Markušić,S.; Ivanˇcić,I. Numerical modelling of the Ston–Slano (Croatia) aftershock sequence. Stud. Geophys. Geod. 2001, 45, 251–266. [CrossRef] 3. Croci, G.; Bonci, A. UNESCO world heritage centre –Mission to Dubrovnik and Ston. Technical Report 1996, 24 pp. Available online: https://whc.unesco.org/document/140421 (accessed on 28 February 2020). 4. Herak, M.; Allegretti, I.; Herak, D.; Kuk, K.; Kuk, V.; Marić,K.; Markušić,S.; Stipˇcević,J. HVSR of ambient noise in Ston (Croatia)—Comparison with theoretical spectra and with the damage distribution after the 1996 Ston–Slano earthquake. Bull. Earthq. Eng. 2010, 8, 483–499. [CrossRef] 5. University of Zagreb, Macroseismic archives of the Department of Geophysics, Faculty of Science, University of Zagreb, Zagreb, 2020; University of Zagreb: Zagreb, Croatia, 2020. 6. Schmid, S.M.; Bernoulli, D.; Fügenschuh, B.; Matenco, L.; Schefer, S.; Schuster, R.; Tischler, M.; Ustaszewski, K. The Alpine–Carpathian–Dinaridic orogenic system: Correlation and evolution of tectonic units. Swiss J. Geosci. 2008, 101, 139–183. [CrossRef] 7. Ustaszewski, K.; Schmid, S.M.; Lugović,B.; Schuster, R.; Schlategger, U.; Bernoulli, D.; Hottinger, L.; Kounov, A.; Fugenschuh, B.; Schefer, S. Late Cretaceous intra-oceanic magmatism in the internal Dinarides (northern Bosnia and Herzegovina): Implication for the collision of the Adriatic and European Plate. Lithos 2009, 108, 106–125. [CrossRef] 8. Vlahović,I.; Tišljar, J.; Velić,I.; Matiˇcec,D. Evolution of the Adriatic Carbonate Platform: Palaeogeography, main events and depositional dynamics. Palaeogeogr. Palaeoclimatol. Palaeoecol. 2005, 220, 333–360. [CrossRef] 9. Schmid, S.M.; Fügenschuh, B.; Kounov, A.; Matenco, L.; Nievergelt, P.; Oberhänsli, R.; Pleuger, J.; Schefer, S.; Schuster, R.; Tomljenović,B.; et al. Tectonic units of the alpine collision zone between Eastern Alps and western Turkey. Gondwana Res. 2020.[CrossRef] 10. Raić,V.; Papeš, J.; Ahac, A.; Korolija, B.; Borović,I.; Grimani, I.; Marinˇcić,S. Basic Geological Map of the SFRY, 1:100,000 Scale, Ston Sheet K33-48. Geoinženjering OOUR Institut za geologiju Sarajevo (1972–1980) and Institut za geološka istraživanja Zagreb (1967–1968); Federal Geological Institute: Beograd, Serbia, 1980. 11. Natević,L.J.; Petrović,V. Basic Geological Map of the SFRY, 1:100,000 Scale, Trebinje Sheet K34-37. Geološki zavod Sarajevo (1963); Federal Geological Institute: Beograd, Serbia, 1967. Remote Sens. 2020, 12, 1157 22 of 23

12. D’Agostino, N.; Avallone, A.; Cheloni, D.; D’Anastasio, E.; Mantenuto, S.; Selvaggi, G. Active tectonics of the Adriatic region from GPS and earthquake slip vectors. J. Geophys. Res. B–Solid Earth 2008, 113, B12413. [CrossRef] 13. Bennett, R.A.; Hreinsdóttir, S.; Buble, G.; Bašić,T.; Baˇcić,Z.; Marjanović,M.; Casale, G.; Gendaszek, A.; Cowan, D. Eocene to present subduction of southern Adria mantle lithosphere beneath the Dinarides. Geology 2008, 35, 3–6. [CrossRef] 14. Ivanˇcić,I.; Herak, D.; Markušić,S.; Sović,I.; Herak, M. Seismicity of Croatia in the period 1997–2001. Geofizika 2001, 18–19, 17–29. 15. Kastelic, V.; Vannoli, P.; Burrato, P.; Fracassi, U.; Tiberti, M.M.; Valensise, G. Seismogenic sources in Adriatic Domain. Mar. Pet. Geol. 2013, 42, 181–213. [CrossRef] 16. Heidbach, O.; Custodio, S.; Kingdon, A.; Mariucci, M.T.; Montone, P.; Müller, B.; Pierdominici, S.; Rajabi, M.; Reinecker, J.; Reiter, K.; et al. Stress map of the Mediterranean and Central Europe 2016. GFZ Data Serv. 2016. [CrossRef] 17. Handy, M.; Ustaszewski, K.; Kissling, E. Reconstructing the Alps–Carpathian–Dinarides as a key to understanding switches in subduction polarity, slab gaps and surface motion. Int. J. Earth. Sci. (Geol. Rundsch.) 2014.[CrossRef] 18. Herak, D.; Herak, M. Focal depth distribution in the Dinara Mt. region, Yugoslavia. Gerlands Beitr. Geophys. 1990, 99, 505–511. 19. Tomljenović,B.; Herak, M.; Herak, D.; Kralj, K.; Prelogović,E.; Bostjanˇcić,I.; Matoš, B. Active tectonics, seismicity and seismogenic sources of the Adriatic coastal and offshore region of Croatia. In 28 Convegno Nazionale "Riassunti Estesi delle Comunicazioni"; Stella Arti Grafice: Trieste, Italy, 2009; pp. 133–136. 20. Herak, M.; Herak, D.; Markušić,S. Revision of the earthquake catalogue and seismicity of Croatia, 1908–1992. Terra Nova 1996, 8, 86–94. [CrossRef] 21. ISC. Bulletin of the International Seismological Centre. Available online: http://www.isc.ac.uk/iscbulletin/ search/ (accessed on 28 February 2020). 22. EIDA. European Integrated Data Archive. Available online: http://eida.gfz-potsdam.de/webdc3/ (accessed on 28 February 2020). 23. IRIS. Incorporated Research Institutions for Seismology. Available online: http://ds.iris.edu/wilber3/find_ event (accessed on 28 February 2020). 24. Molinari, I.; Morelli, A. EPcrust: A reference crustal model for the European plate. Geophys. J. Int. 2011, 185, 352–364. [CrossRef] 25. Stipˇcević,J.; Herak, M.; Molinari, I.; Dasović,I.; Tkalˇcić,H.; Gosar, A. AlpArray-CASE Working Group and AlpArray Working Group, crustal thickness beneath the Dinarides and surrounding areas from receiver functions. Tectonics 2020, 39, e2019TC005872. [CrossRef] 26. Herak, M. Hyposearch: An earthquake location program. Comput. Geosci. 1989, 15, 1157–1162. [CrossRef] 27. Herak, M.; Herak, D. Seismicity and seismic sources in the greater Kvarner–Velebit area, Croatia: Recent advances. In Proceedings of the 5th Regional scientific meeting on quaternary geology dedicated to geohazards, Starigrad-Paklenica, Croatia, 9–10 November 2017; pp. 20–21. 28. Yang, X.; Bondár, I.; McLaughlin, K.; North, R. Source specific station corrections for regional phases at Fennoscandian stations. Pure Appl. Geophys. 2001, 158, 35–57. [CrossRef] 29. Materni, V.; Giuntini, A.; Chiappini, S.; Chiappini, M. Relocation of earthquakes by source-specific station corrections in Iran. Bull. Seismol. Soc. Am. 2015, 105, 2498–2509. [CrossRef] 30. Dziewonski, A.M.; Chou, T.A.; Woodhouse, J.H. Determination of earthquake source parameters from waveform data for studies of global and regional seismicity. J. Geophys. Res. 1981, 86, 2825–2852. [CrossRef] 31. Ekström, G.; Nettles, M.; Dziewonski, A.M. The global CMT project 2004–2010, Centroid-moment tensors for 13,017 earthquakes. Phys. Earth Planet. Inter. 2012, 200–201, 1–9. [CrossRef] 32. USGS. U.S. Geological Survey Earthquake Hazards Program. Available online: https: //earthquake.usgs.gov/earthquakes/search/https://earthquake.usgs.gov/earthquakes/search/https: //earthquake.usgs.gov/earthquakes/search/ (accessed on 28 February 2020). 33. GCMT. Global Centroid Moment Tensor Catalog. Available online: https://www.globalcmt.org/CMTsearch. html (accessed on 28 February 2020). 34. ISC. International Seismological Centre Global Focal Mechanism Catalogue. Available online: http: //www.isc.ac.uk/iscbulletin/search/fmechanisms/ (accessed on 28 February 2020). Remote Sens. 2020, 12, 1157 23 of 23

35. ISC-GEM. International Seismological Centre Global Instrumental Earthquake Catalogue, 2020. Available online: https://doi.org/10.31905/d808b825 (accessed on 28 February 2020). 36. RCMT. European-Mediterranean Regional Centroid-Moment Tensors. Available online: http://rcmt2.bo.ingv. it/Italydataset1976-2015.dek (accessed on 28 February 2020). 37. Musson, R.M.W. MEEP 2.0 User Guide; Open Report OR/09/045; British Geological Survey: Keyworth, Nottingham, UK, 2009; p. 16. 38. Herak, M.; Živˇcić,M.; Sović,I.; Cecić,I.; Dasović,I.; Stipˇcević,J.; Herak, D. Historical seismicity of the Rijeka region (NW External Dinarides, Croatia)–Part II: The Klana earthquakes of 1870. Seismol. Res. Lett. 2018, 89, 1524–1536. [CrossRef] 39. Herak, D.; Živˇcić,M.; Vrkić,I.; Herak, M. The Međimurje (Croatia) earthquake of 1738. Seismol. Res. Lett. 2020, 91, 1042–1056. [CrossRef] 40. 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] 41. Rosen, P.A.; Gurrola, E.; Sacco, G.F.; Zebker, H. The InSAR scientific computing environment. In Proceedings of the EUSAR. 9th European Conference on Synthetic Aperture Radar, Nuremberg, Germany, 23–26 April 2012; pp. 730–733. 42. Farr, T.G.; Rosen, P.A.; Caro, E.; Crippen, R.; Duren, R.; Hensley, S.; Kobrick, M.; Paller, M.; Rodriguez, E.; Roth, L.; et al. The shuttle radar topography mission. Rev. Geophys. 2007, 45, RG2004. [CrossRef] 43. Massonnet, D.; Feigl, K.L. Discrimination of geophysical phenomena in satellite radar interferograms. Geophys. Res. Lett. 1995, 22, 1537–1540. [CrossRef] 44. Hanssen, R.F. Radar Interferometry: Data Interpretation and Error Analysis; Springer Science & Business Media: Berlin/Heidelberg, Germany, 2001; Volume 2. 45. Scharoo, R.; Visser, P.N.A.M.; Mets, G.J. Precise orbit determination and gravity field improvement for the ERS satellites. J. Geophys. Res. 1998, 88, 4984–4996. [CrossRef] 46. Goldstein, R.M.; Werner, C.L. Radar interferogram filtering for geophysical applications. Geophys. Res. Lett. 1998, 25, 4035–4038. [CrossRef] 47. Chen, C.W.; Zebker, H.A. Two-dimensional phase unwrapping with use of statistical models for cost functions in nonlinear optimization. J. Opt. Soc. Am. 2001, 18, 338–351. [CrossRef] 48. Ortner, H.; Reiter, F.; Acs, P. Easy handling of tectonic data: The programs Tectonic VB for Mac and Tectonics FP for Windows TM. Comput. Geosci. 2002, 28, 1193–1200. [CrossRef] 49. Turner, F.J. Nature and dynamic interpretation of deformation lamellae in calcite of three marbles. Am. J. Sci. 1953, 251, 276–298. [CrossRef] 50. Marrett, R.; Allmendinger, R.W. Kinematic analysis of fault-slip data. J. Struct. Geol. 1990, 12, 973–986. [CrossRef] 51. Angelier, J.; Mechler, P. Sur une méthode graphique de recherche des contraintes principales également utilisable en tectonique et en séismologie: La méthode des dièdres droits. Bull. Soc. Géol. Fr. 1977, 19, 1309–1318. [CrossRef] 52. Wells, D.L.; Coppersmith, K.J. New empirical relationships among magnitude, rupture length, rupture width, rupture area, and surface displacement. Bull. Seismol. Soc. Am. 1994, 84, 974–1002. 53. Funning, G.J.; Garcia, A. A systematic study of earthquake detectability using Sentinel-1 Interferometric Wide-Swath data. Geophys. J. Int. 2019, 216, 332–349. [CrossRef] 54. Ross, Z.E.; Idini, B.; Jia, Z.; Stephenson, O.L.; Zhong, M.; Wang, X.; Zhan, Z.; Simons, M.; Fielding, E.J.; Yun, S.-H.; et al. Hierarchical interlocked orthogonal faulting in the 2019 Ridgecrest earthquake sequence. Science 2019, 366, 346–351. [CrossRef] 55. Chounet, M.; Vallée, M.; Causse, M.; Courboulex, F. Global catalog of earthquake rupture velocities shows anticorrelation between stress drop and rupture velocity. Tectonophysics 2018, 733, 148–158. [CrossRef]

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).