1 Introduction

South Evia Island is a low seismicity region of central Greece where no onshore strong earthquake record exists (Fig. 1; Papazachos and Papazachou 1997; Goldsworthy et al. 2002). An instrumental seismicity catalog with earthquakes of local magnitude M4+ for the period 1963–2003 shows no events ML> 5.0 have occurred within a 50-km radius from South Evia (see supplementary Figure S1a and Table S1; data source: NOA). Evia is characterized by mountainous topography with a general NW-SE orientation, parallel to the strike of large normal faults that accommodate present day extension (Roberts and Jackson 1991; Roberts and Ganas 2000). The active normal faults control the geomorphology and they bound two Plio-Quaternary marine grabens (North and South Gulf of Evia; Fig. 1) whose asymmetry is evident in the topography, bathymetry, vertical movements of the coastline and tilting of syn-rift sediments adjacent to them (e.g., Roberts and Jackson 1991; Perissoratis and Van Andel 1991; Pirazzoli et al. 1999; Cundy et al. 2010; Evelpidou et al. 2011; Valkanou et al. 2021; Caroir et al. 2023). Onshore syn-rift sequences are of Upper Miocene–Pliocene–Quaternary age (Mettos et al. 1991; Rondoyanni et al. 2007; Palyvos et al. 2006). The south part of the island is built with metamorphic bedrock (mainly schists and marbles; IGME 1991), carbonates, and a small Miocene-Pliocene basin (Kymi-Aliveri basin; Kokkalas 2001). While no large faults have been mapped onshore, it is well known that several active faults occur offshore, inside the South Gulf of Evia (Papanikolaou et al. 1988; Perissoratis and Van Andel 1991) and on the Aegean side, south of the Island of Skyros (e.g., Stiros et al. 1992).

Fig, 1
figure 1

Relief map of South Gulf of Evia region, central Greece. Beach balls indicate fault plane solutions of moderate seismic sequences since 1999 (red colors; source NOA). With blue beach balls, it is shown the 29 November 2022 (Mw 4.6–Mw 4.8), 28 December 2022 (Mw 4.7), 4 January 2023 (Mw 4.1), and 22 April 2023 (Mw 4.5) events. Brown lines are active faults from Ganas et al. (2013, 2023). The events with focal mechanism data are reported in Table S2. Focal mechanism for the 18 June 2003 event is from Benetatos et al. (2004). Focal mechanism for the 9 June 2015 event is from Ganas et al. (2016). Black box shows extent of Figs. 2, 3, 4. Inset box shows location of study area within Greece

Full size image

During October 2022–April 2023, a shallow earthquake sequence occurred onshore South Evia Island near the villages Zarakes and Almyropotamos (Fig. 1; Evangelidis and Fountoulakis 2023). The sequence initiated on 25 October 2022 (Fig. 1) with small size events that occurred in the vicinity of the mainshock, south of Almyropotamos (see Fig. 1 black dots) and has been peaked with a shock of magnitude ML= 4.8 (Mw = 4.6) that occurred on 29 November 2022 04:32 UTC (event-1). This event was followed by a stronger ML = 5.0 (Mw = 4.8) seismic event on the same day at 20:06 UTC (event-2), which was felt even in the city of Athens. Its seismic intensity reached V, according to the Institute of Engineering Seismology and Earthquake Engineering (ITSAK) measurements (Fig. S2) while only light damages were recorded in South Evia. During the following weeks many shallow aftershocks occurred, mostly onshore, while on 22 April 2023 08:38, UTC another moderate event ML = 4.5 (Mw = 4.5; Fig. 1) occurred as well. Despite the moderate magnitude of the earthquakes and while intensity did not exceed V (Fig. S2), the sequence appeared to be interesting (from a seismotectonic point of view), driven by the fact that it took place in an area where no mapped fault existed on literature maps. Besides, previous seismic activity includes only a seismic swarm that occurred during 2003, at about 60 km to the north-west of the 2022–2023 sequence, near Psachna (Fig. 1; Benetatos et al. 2004; Papoulia et al. 2006). The background seismicity (considering shallow events with M2.0+) in both central and South Evia includes several hundreds of events scattered both onshore and offshore Evia (Fig. S1b, Fig. S1c; source https://bbnet.gein.noa.gr/HL/databases/database). Near the epicenters of the 2022 sequence, the occurrence of background events is mostly offshore, inside the South Gulf of Evia, as well as in NE Attica (see Fig. S1b).

In this study, we use regional seismological data to identify and characterize the seismic fault (location, kinematics, depth range) and its potential. The identification of an unknown fault is of key importance especially since the epicentral area neighbors populated towns of Evia Island and NE Attica but also lays in a distance less than 50km from the city of Athens; therefore, it could affect the regional seismic risk. We also use uniform-slip, near-vertical fault models to investigate the static stress transfer on receiver faults in South Evia and the neighboring crust. We further explore the variation of one strong motion parameter, the peak ground velocity (PGV) to assess the possibility of delayed dynamic triggering of two earthquake sequences, which took place at about 60 km to the NW, onshore central Evia (Fig. 1) respectively. The two latter sequences occurred in December 2022 and January 2023, motivating us to explore possible triggering relationships.

2 Tectonic setting

The island of Evia (also written as Euboea; Fig. 1) is located between the central Aegean region to the east and Attica–Viotia regions to the west (Fig. 1). The central Aegean mainly hosts fault segments of the right-lateral North Anatolian fault system and conjugate left-lateral faults (Papadopoulos et al. 2002; Karakostas et al. 2003; Ganas et al. 2005; Chatzipetros et al. 2013; Ganas et al. 2014; Karakostas et al. 2014; Papanikolaou et al. 2019). Attica forms the south margin of the E-W Quaternary rift of South Viotia–Oropos (Ambraseys and Jackson 1990; Goldsworthy et al. 2002; Ganas et al. 2005; Chousianitis et al. 2013; Grützner et al. 2016; Briole et al. 2021; Valkaniotis et al. 2023). Along the north coast of Attica, a set of parallel, north-dipping large normal faults accommodate NE-SW extension (e.g., IGME 1991; Grützner et al. 2016; Deligiannakis et al. 2018; Iezzi et al. 2021) which is also evidenced by Global Navigation Satellite System (GNSS) data (Chousianitis et al. 2013; Chousianitis et al. 2015; D'Agostino et al. 2020; Briole et al. 2021) and focal mechanism data (e.g., Kaviris et al. 2018; Konstantinou et al. 2020). These coastal normal faults of Attica face an opposing series of south-dipping normal faults that have been mapped along the southern coastline of Evia (Perissoratis and Van Andel 1991). The offshore area (i.e., between Attica and South Evia) is seismically active in the depth range 5–20 km (Papoulia et al. 2006; Konstantinou et al. 2020). Further evidence for a full-graben structure regarding the South Gulf of Evia rift has been suggested by Rondoyanni et al. (2007) for the region to the SE of town of Halkida (Fig. 1).

Further northwest, an active deformation component of left-lateral shear was identified for the first time inside the North Gulf of Evia rift from seismological data (Ganas et al. 2016). The seismological data provided evidence that on 9 June 2015 event 01:09 UTC a left-lateral, NW-SE striking, near-vertical fault, ruptured at mid-crustal depths (~10-km) about 15 km NW of Halkida (Fig. 1). The left-lateral shear faulting inside the Gulf of Evia is kinematically compatible to the ongoing NNE-SSW extension (N14°E; Roberts and Ganas 2000) accommodated by moderate to high-angle normal faults. In addition, inside the north gulf of Evia several E-W to NW-SE minor normal faults exist (Sakellariou et al. 2007; Karastathis et al. 2007; Caroir et al. 2023) that are interpreted as secondary structures with moderate seismic potential. As both rifts (North and South Gulf of Evia) are of similar Plio-Quaternary age (Roberts and Jackson 1991; Roberts and Ganas 2000; Goldsworthy et al. 2002), it may be expected that similar, left-lateral shear faults occur at mid-crustal depths, with moderate seismic potential.

Across the south gulf of Evia estimates of tectonic strain from Global Navigation Satellite System (GNSS) data (Chousianitis et al. 2013; Chousianitis et al. 2015; D'Agostino et al. 2020) indicate NNE-SSW extension. The rate of extension ranges from 1.2 mm/year (equivalent to 1-D strain rate of23.2 ns/year) to 0.6 mm/year (13 ns/year) toward the southern termination of the rift. While the main component of the deformation is extensional, the occurrence of left-lateral shear inside the North Gulf of Evia (Ganas et al. 2016) points out that a similar slip component may be active inside the South Gulf. The occurrence of the 2022–2023 sequence gave us the data necessary to examine the kinematics of the mainshock and identify its nature.

3 Data and methods

3.1 Relocation of seismicity

We study the seismicity sequence in South Evia (Central Greece) between the 25thof October 2022, when the first moderate events occurred, and the 23rd of April 2023. The sequence was recorded by the broad-band and strong motion stations of the regional Hellenic Unified Seismic Network (HUSN) stations and broad-band stations of the AdriaArray project (a new European geophysical initiative; Kolínský et al. 2022), as well (Fig. S3). P- and S- arrival times used in this study were manually picked by the Geodynamics Institute of the National Observatory of Athens (GI-NOA).

The sequence includes about 800 seismic events with magnitude ML between 0.5 and 5.0, which were initially located using the HYPOINVERSE code (Klein 2002). Several velocity models were examined during the location–relocation procedure such as Konstantinou (2018), Konstantinou et al. (2020), Mouzakiotis and Karastathis (2016) and Kaviris et al. (2007) (Fig. S4). The comparison was initially performed on the HYPOINVERSE results including the determination of location errors (Fig. S5) and the epicenters’ distribution (Fig. S4). The crustal model of Kaviris et al. (2007) was finally selected since it depicted the lowest errors and has been derived from a recent seismic experiment including the district of the study area (Fig. S5). The average absolute location errors reported by HYPOINVERSE where mean rms (root mean square) travel time error 0.14 s, ERH=0.61 km, and ERZ=0.95 km for horizontal and vertical uncertainties, respectively. The value of Vp/Vs ratio was set to 1.75 according to the Wadati diagram results (Fig. S6). The applied distance weighting was set to nullify the influence of stations at epicentral distances further than ~120 km.

In order to improve the initial events’ locations, the double difference relocation HypoDD (Waldhauser 2001) procedure was performed. HypoDD determines relative locations within clusters, using a double difference algorithm, developed by Waldhauser and Ellsworth (2000). It improves relative location accuracy by strongly reducing the influence of the velocity structure on locations and uncertainties caused by arrival time readings errors, by minimizing the double-difference between observed and calculated travel times through iterative weighted least squares using the conjugate gradients method. The relocation procedure takes into account both waveform cross-correlation and catalog differential travel time data to provide an enhanced picture of the seismic activity.

Waveform cross-correlation was incorporated in the relocation procedure, for the closest 21 stations (distance <75km, Fig. S3), in order to enhance the strong correlated events by decrease of the relative location uncertainties induced by the P- and S-wave arrivals picking inaccuracies. Full signal waveforms were cropped in windows containing both P and S waves, filtered between 2 and 6 Hz, and a threshold of 70% was set concerning the waveform coherency. The double-difference residuals for the pairs of earthquakes at each station were minimized by weighted least squares, using the method of conjugate gradient least squares (LSQR). Errors reported by LSQR are grossly underestimated and need to be assessed independently by using the singular value decomposition technique (SVD) on a subset of events (Waldhauser and Ellsworth 2000). Therefore, in a first step, a subset of the sequence, including the strongest events (ML>3), was relocated using the SVD method (Fig. S7) indicating onshore activity-oriented NW-SE. A good data fit of mean rms< 0.12s was obtained. Next, the LSQR scheme was followed for the total dataset including data from stations in a maximum distance of 75km (Fig. S3; this distance includes SKY station to the North to reduce the azimuthal gap). The velocity model used in the relocation was the model used in the initial location process. Variations in station distribution for each event pair can introduce errors in relative event locations that cannot be quantified directly (Waldhauser and Ellsworth 2000). We apply the jack-knife method (e.g., Efron 1982) to estimate the variance of errors in each coordinate direction. The procedure involves repeated relocation, each time sub-sampling the data by deleting one station at a time. No outliers are removed during this process. The mean and median absolute perturbation on each direction (Table 1) and the 95% confidence interval ellipsoids (Fig. S8) are then determined from the distribution of the perturbed hypocenter positions for each event. The majority of the events are described by a small error ellipsoid (less than 1 km; Fig. S8) which indicates that the relevant relocated positions are robust. The HypoDD final results include 88% of the initial dataset (701 events) (rms residual for cross-correlated data: 0.22 s, rms residual for catalog data: 0.16 s, Fig. S9).

Table 1 Statistics from the station jack-knife test of relocation robustness
Full size table

3.2 Focal mechanisms

The moment tensor (MT) solutions of the strongest events were calculated by GI-NOA (https://bbnet.gein.noa.gr/HL/seismicity/mts), yet to confirm the source of the weaker events the MT inversion was performed for the three weaker aftershocks (Mw between 3.5 and 3.9) of the sequence (Table 2, Fig. 2). The ISOLA platform was employed (for detailed description see Sokos and Zahradnik 2008; Zahradník and Sokos 2018). Complete waveforms were used, without separation of individual phases; full-wave Green’s functions were calculated by the discrete wavenumber method (Bouchon 1981). The 1D velocity model suggested by Kaviris et al. (2007) and the ratio Vp/Vs value 1.75 were adopted. The inversion was performed in the frequency bands 0.04–0.09 and 0.06–0.12Hz, depending on the magnitude of the event and the distance from the station used.

Table 2 Moment tensor solutions calculated in this study and by GI-NOA (*)
Full size table
Fig. 2
figure 2

a Map showing the spatial distribution of the relocated (HypoDD) sequence (see also Fig. S10a for a plot with transparent, depth-free, events). The MT solutions (Table 2) are plotted as beach balls (The shaded quadrants indicate compression and the light quadrants indicate extension). b, c, d Distribution of the relocated (HypoDD) sequence at cross-sections A1-A2 (b) and B1-B2 (c), corresponding to Fig. 2.a. Notice the near-vertical structure (about 2 km wide) that is visible in the left panel (i.e., across strike). The length of the structure is about 6 km (middle panel). d Additional projection of the cross-section B1-B2, where green and blue colors indicate the foreshocks and aftershocks of event-1 (MT is shown), respectively

Full size image

The results are expressed in terms of the double-couple component of the deviatoric solution, represented by the scalar moment, strike, dip, and rake. Several HUSN network broadband stations (ATH, SKY, THVA, MRKA, KARY) were considered in the process. The variance reduction (VR) was >60% in all three cases, while the double-couple (DC) percentage was >70% (Fig. S11). The quantification of the focal mechanism variability in FMVAR is done via the use of Kagan angle (i.e., the minimum rotation angle between two P-T-N coordinate systems describing the focal mechanisms under comparison). The angle can vary from 0° (perfect agreement between two solutions) to 120° (total disagreement), the values below 20°–40° indicate a good agreement. For more details and references, see paper by Sokos and Zahradnik (2013). In our study, FMVAR values were between 12° and 18°.

3.3 Parameters of the earthquake sequence

The seismic sequence debuted in Evia Island on October 25 and 26, 2022, and was followed by 55 events until late November 2022 (Fig. 4). The seismicity seemed to pause until November 28, when limited number of events took place, while during the November 29th, the two major events (M4.6 & M4.8) occurred. Throughout the next days, the seismic sequence showed a temporal decay until the 3rd of December 2022, when a stage with several events of magnitude ML>3.0 initiated (Figs. 3, 4). Next, on April 22, 2023, another moderate event (Mw=4.5) occurred, followed by several weak events while during the next days the seismicity demonstrated some degradation.

Fig. 3
figure 3

Map of South Evia showing the time distribution of the relocated sequence events colored according to time of occurrence for the time period 25 October 2022–23April 2023. Date on color scale is formed as DDMMYY. Thick black line indicates line of projection of events shown in Fig. 4

Full size image
Fig. 4
figure 4

Space-time evolution of the South Evia sequence projected along a 10-km section AB (thick black line in Fig. 3), for the time period 25 October 2022–23 April 2023. The color scale is the same as in Fig. 3

Full size image

The Z-map package (Wiemer 2001) has been applied for obtaining the characteristics of the sequence such as the cumulative number of events with time (Fig. 5). We also studied the frequency–magnitude distribution of the 2022–2023 earthquakes. The estimation of the completeness magnitude (Mc) is a required step for any seismic sequence study. This magnitude, Mc, is the minimum magnitude above which all events are reliably detected in a defined region and time. This is crucial for further seismicity analysis, given that a magnitude lower than Mc leads to biased results (Rydelek and Sacks 1989; Kagan 2004; Mignan et al. 2011; Olasoglou et al. 2016). Wiemer and Wyss (2000) assessed the minimum magnitude for complete earthquake data. Then, parameters a, b were computed with a Maximum likelihood solution method for 55 events (foreshocks) of the period 25.10.2022–28.11.2022 (blue color) and the main sequence period, 29.11.2022-23.4.2023 (green color); the graph is depicted in Fig. 6.

Fig. 5
figure 5

Frequency-time evolution of the sequence for the time period 25 October 2022–24 May 2023 (extended period is plotted to demonstrate the sequence decay). Yellow stars denote the major events with ML>4.0

Full size image
Fig. 6
figure 6

Frequency–magnitude graph showing maximum likelihood solution for the seismic sequence period 29.11.2022–23.4.2023 (green color) b-value=0.78(+/−0.05), and the foreshock sequence period 25.10.2022–28.11.2022 (blue color) with b=1.37 (±0.25). The magnitude of completeness Mc = 1.8 is the same in both cases

Full size image

4 Earthquake interaction

4.1 Static triggering and relation to aftershock patterns

It is well established that the coseismic slip along the fault plane causes changes in static stress patterns that may trigger subsequent earthquakes on neighboring segments as well as aseismic slip on unruptured patches of the seismic fault itself (e.g., Lin and Stein 2004; Ganas et al. 2006; Atzori et al. 2008; D’Agostino et al. 2012; Kassaras et al. 2022). In this context, it is important for hazard assessment to identify other possible faults in the broader Evia region that may have been brought closer to failure by the stress changes associated with the 2022 South Evia earthquakes. To investigate the Coulomb stress variations, we calculated the Coulomb stress change (Coulomb Failure Function, CFF, or Coulomb stress; Reasenberg and Simpson 1992), induced by the mainshock using our fault model (Table 3). We assume a rectangular dislocation model with a uniform-slip size of 3 by 3 km constrained by the cross-sections of the relocated seismicity, including the extent of aftershocks (Fig. 2). We consider as source for the M4.6 strike-slip event-1 the dimension of a single slip patch, provided by the empirical relationship of Wells and Coppersmith (1994): log RA = a + b × M (where a = −3.42 and b=0.9; RA is rupture area) for strike slip faults. This formula provides a RA= 5.24 km2 for a M4.6 source and 7.94 km2 for a M4.8 source (event-2), respectively. Given the uncertainties associated with regression statistics and the moderate size of the events, we considered as source dimension a square patch of 3 × 3 km (or 9 km2). A 3-km size for fault length seems a suitable approximation given that the distance between the epicenters of the two events is less than 2 km (Fig. 2) while the hypocentral depths are at 8 km (event-2) and 9 km (event-1).

Table 3 Parameters of fault models for event-1 (29 November 2022 04:32 UTC) and event-2 (29 November 2022 20:29 UTC) used in Coulomb stress modeling, respectively
Full size table

First, we calculate the Coulomb stress change due to rupture on event-1 on fault segments with the same kinematics and geometry (Fig. 7) at the depth range 8–9 km, and assuming an effective friction coefficient μ′=0.4, as is commonly used in stress interaction studies (e.g., Harris and Simpson 1998; Freed 2005). The Coulomb stress change was computed in an elastic half space (Okada 1992) by assuming a shear modulus of 3.0 × 1010 Pa, Poisson’s ratio 0.25. The parameters of the two fault models are given in Table 3 using the MT determinations of this study (Table 2).

Fig. 7
figure 7

Map of Coulomb stress computed on left-lateral receiver faults striking N134°E due to event-1, at 9-km depth (a) (see Fig. S12 for a stress cross-section at right-angles to the causative fault.) and 8-km depth (b). Color palette of stress values is linear in the range – 2 to + 2 bar (1 bar=100 kPa). Blue areas indicate unloading, red areas indicate loading, respectively. Black color shows the area where stress reduction was < −2 bar. Green circles are aftershocks between 04:32 and 20:29 UTC. Note that the aftershocks are aligned NW-SE parallel to the strike of the 1st rupture (yellow line)

Full size image

The Coulomb stress change patterns form both along- (i.e., N134°E) and across-strike positive stress lobes (red color in Fig. 7); however, most of the event-1 aftershocks occur in negative (relaxed) stress areas (or stress shadows, Harris and Simpson 1998; see a cross-section orthogonal to fault’s strike in Fig. S11 for a depth distribution). This is probably caused by the assumption that coseismic slip was uniformally distributed along the fault planeand/or that some of off-fault aftershocks occurred on fault planes exhibiting different kinematics than those of event-1. However, most of the event-1 aftershocks are oriented NW-SE in map-view (i.e., similar orientation to the modeled fault plane) and they are located ± 1 km from the fault (Fig. 7) so it is reasonable to assume that they may be co-planar to event-1 (i.e., on-fault aftershocks). The hypocenter of event-2 is located about 2 km to the NW of event-1 at a depth of 8 km; therefore, it may have been triggered by Coulomb stress tranfer as expected; however, it is not confirmed because of the modeling limitations (the uniform slip assumption includes the hypocentral area of event-2; Fig. 7b).

Next, we calculated the stress changes due to event-2 (29 November 2022 20:29 UTC; Mw=4.8). At 8-km depth the lobe pattern is similar to that of event-1 except for the amplitudes of transferred stress that are larger (Fig. 8). Two large lobes with negative stress levels (stress shadows) have formed in a N-S direction. Most aftershocks of both events occurred to the southwest of event-2 epicenter, while many events occurred inside the stress shadows. Several active faults inside the South Gulf ofEvia were loaded with Coulomb stress less than 0.1 bar. The impact of the stress loading on target planes diminishes after a distance of 7 km from the epicenter of event-2, where the static stress falls below +0.1 bar. The events of the 22–23 April 2023 seismic burst occurred on either side of the M4.8 event (yellow circles in Fig. 8). The M4.5 aftershock occurred about 1.5km NW of the M4.8 event (event-2).

Fig. 8
figure 8

a Map of Coulomb stress computed on left-lateral receiver faults striking N134°E due to event-2, at 8-km depth. b Map of Coulomb stress computed on optimally oriented faults to regional extension N18°E due to event-2, at 8-km depth. Color palette of stress values is linear in the range − 1 to + 1 bar (1 bar=100 kPa). Blue areas indicate unloading, red areas indicate loading, respectively. Black color shows the area where stress reduction was < −1 bar and white color areas where stress increase was > 1 bar. Green circles are aftershocks between 29 November 20:29 UTC and 28 December 2022.Red line indicates the surface projection of the modeled, left-lateral seismic source. Yellow circles represent events of the April 22–23 seismic burst

Full size image

We also modeled static stress transfer patterns due to event-2 resolved on optimal planes to regional extension. We used the extension direction found by Chousianitis et al. (2015) based on GNSS data, that is N18°E for this area of South Evia (we used the grid node located 13 km to the NE of the 2022 events). A similar strain axis orientation (NNE-SSW) was obtained by D'Agostino et al. (2020) and Konstantinou et al. (2020). The stress pattern (at 8-km depth) shows positive lobes to the SE of the epicenteraWS of event-2 where many afteshocks are located (Fig. 8). The loading of this area reached levels greater than +2 bar (Fig. 8) and this may justify the occurrence of aftershocks forseveral months (end of April 2023; see time plot in Fig. 4). We attribute the occurrence of off-fault aftershocks inside the stress shadows to the uniform-slip assumption along the fault plane, used also in the CFF modeling due to 2nd rupture. Nevertheless, the uniform slip model may explain the overall NW-SE arrangement of off-fault aftershocks (Fig. 8). Similarly to the specified target faults case, the static stress loading falls below 0.1 bar at distances larger than 7 km from the mainshock.

5 Discussion

5.1 Seismotectonic setting and left-lateral shear onshore South Evia

In this section, we discuss our seismological findings from a very comprehensive investigation of the source of 29 November 2022 Mw 4.6 and 4.8 earthquakes along a blind, strike-slip fault onshore South Evia Island. This seismic sequence occurred in a critical area, a few km to the east of a marine graben (i.e., South Gulf of Evia), which hosts active faults capable of generating potentially large (M > 6) earthquakes (i.e., Goldsworthy et al. 2002; Grützner et al. 2016; Deligiannakis et al. 2018; Iezzi et al. 2021).

The majority of the relocated events of 5-months long period, are distributed in a dense cluster, onshore South Evia (Fig. 2.a, Fig. S10a) indicating the activation of the NW-SE striking fault plane at shallow depths (h< 12 km, Fig. 2.a, b). The spatial distribution shown in the cross sections A1-A2 (across-strike) and B1-B2 (along strike; Fig. 2.b, c) reveals that the relocated hypocenters form an almost vertical structure, in depth between 6 and 11.5km. The structure appears coherent with focal mechanisms (Table 2) of the major events (strike N134°E, dip 88°). Furthermore, we applied the HC method (Zahradnik et al. 2008) in order to identify which of the nodal planes, obtained by the MT solution, is the causative fault (Fig. S10b and c); the results support the NW-SE nodal plane. Moreover, the additional solutions reveal that even the weaker events, which occurred a few days later than the 29 November events, are also attributed to strike—slip movement source like the stronger events. As indicated by the relocated seismicity data, there is a single causative fault, while its overall rupture zone dimensions are about 6 × 6 (km2). The overall picture of the relocated aftershocks pattern indicates the activation of a single, left-lateral, near-vertical fault segment hosting the two major shocks of the sequence (see Fig. 2), as well as the 22 April 2023 M4.5 event. Certain events located in a distance, toward NE, may indicate the existence of another structure (described by Evangelidis and Fountoulakis 2023), which is not well illustrated. Finally, few events to the NNW are not connected to the main cluster and may belong to the activation of small faults.

The results of the space–time analysis showed that until the 12th of December 2022, the sequence demonstrated decay, yet after the 14th of December 2022, the seismicity rate shows some recovery (Fig. 5). Moreover, while the sequence evolution until 22 of April 2023 exhibited certain decay, the occurrence of the Mw4.5 event seemed to initiate a new stage. The spatial distribution of the seismic events (Fig. 3) does not reveal any migration of the seismicity, to a different epicentral area in South Evia, although a small number of events are located more distant from the cluster’s center. The evolution of the sequence, as seen in Fig. 5, implies that it was still in progress until 23 April 2023 and aftershocks were still expected; an extended look until May 2023 shows that the seismicity rate tended to stabilize. The overall assessment of this sequence includes a series of shallow events occurring within a rupture zone of about 6-km long (Fig. 2), oriented NW-SE onshore South Evia. Most of the events show an almost uniform distribution except for the last period (2023) where most of the earthquakes have occurred toward the NW part, i.e., near Almyropotamos (Figs. 3, 4). In comparison to Evangelidis and Fountoulakis (2023) findings, we confirm the NW-SE orientation of the causative fault, yet we note that the distribution of epicenters in our study suggests that the seismic fault is mostly located onshore; During the testing of different velocity models (see also Sec. 3.1), we observe significantly higher residuals for the model by Konstantinou et al. (2020) that was used for the relocations in the study by Evangelidis and Fountoulakis (2023). This discrepancy may explain the difference in resulting event locations compared to their study.

Moreover, the statistical analysis shows that the b-value of the foreshocks is notably higher than that of the aftershocks (1.37 vs. 0.78). Despite the relatively low value of the b-parameter of the population of aftershocks (0.78), we note that a comparable b-value=0.86 was obtained by Karakostas et al. (2014) for the aftershock sequence of the 2013 N. Aegean Sea earthquake.

5.2 Foreshock activity patterns

During the seismicity period October–November 2022, we identified 55 events as foreshocks (Fig. 9) including an event ML=3.2 on 18 November 2022 and an ML=3.3 event on 28 November 2022. The foreshock events occurred within 35days before the 29 November 2022 mainshock with local magnitude between 1.4 and 3.3. Most foreshocks (55) occurred within 1-km radius from the epicenter of the 29 November 2022 04:32 UTC event. The foreshock activity started on 25 October 2022, 35 days before the mainshock and fits the pattern of other sequences in Central Greece where such sequences originate within four months from the mainshock (e.g., Papadopoulos et al. 2000). The 2022 foreshocks occurred in a sequence of two groups (25–26 October, 10–22 November), the last one 7 days before the mainshock (see Fig. 4 for a time evolution plot of the events). The occurrence of foreshocks around the hypocenter of event-1 (Fig. 9; Fig. 2.d) favors a cascade-up model (i.e., Ellsworth and Bulut 2018) for the nucleation of event-1, where the sequence of foreshocks builds up both temporally (Fig. 5) and spatially (Fig. 2d and Fig. 9) around the hypocenter of the forthcoming mainshock. Moreover, by examining the catalog ~1.0h before the ML= 4.8 event, we confirm that six (6) immediate foreshocks were observed. The foreshocks also appear to fit better a NW-SE arrangement although the scatter is considerable. Furthermore, the frequency-magnitude distribution of the foreshocks is notably different than the post-mainshock sequence as the foreshock b-value = 1.37 vs. 0.78 (see plot in Fig. 6). The relatively large b-value of the foreshock sequence may indicate relatively low differential stress level in the upper crust (Scholz 1968; see Fig. 4 for depth distribution), or it may be related to high fluid pressure associated with geothermal fluids that are known to exist at shallow levels in South, central and North Evia (e.g., Karastathis et al. 2011; Kanellopoulos et al. 2016). The distribution of the geothermal fields near Halkida (Fig. 1; Lilantio Pedio) is described by Kavouridis and Papadeas (1990), Hurter and Schellschmidt (2003), and Karytsas et al. (2019), while Tombros et al. (2021) studied the association between fault orientation and the flow of the hydrothermal fluids in Kallianos (South Evia) area. We thus documented that a clear foreshock activity existed before the two moderate, intraplate events.

Fig. 9
figure 9

Map showing relocated epicenters of the 2022 South Evia sequence including foreshocks (black dots). The yellow circle indicates 1-km distance from the epicenter of event-1. Thick yellow line indicates surface projection of the seismic fault. Beach balls indicate focal mechanisms of main events (lower hemisphere projections)

Full size image

5.3 Dynamic triggering of the central Evia events

After the earthquakes of November 2022, two short sequences were initiated about 60-km to the NW in central Evia (Fig. 10; Psachna and Vlahia regions). The sequence A started with a ML=4.9 event (the major one) on 28 December 2022 and lasted until January 2023, while the sequence B began on 4 January 2023 with a ML=4.2 event (the major one) and lasted until the end of January (Fig. 10, Fig. S13). These events occurred over a distance several times the dimensions of the 29 November 2022 mainshock fault rupture (3 by 3 km; Fig. 9). The central Evia (Psachna; area A) sequence occurred inside a region which was nearly dormant for 20 years, since June 2003 (see Fig.1). Inspection of the seismic catalogs (Table S1 and Fig. S1a, b, c) reveals that there were not any events ML>=4.3 located in areas A and B from July 2003 to December 2022. We suggest that this timing may not be coincidental but that the two sequences of central Evia are related to the main one of the South Evia.

Fig. 10
figure 10

Relief Map of south and central Evia showing the distribution of the events which occurred in triggered areas A and B, two months before (blue circles) and two months after (A, yellow and B, magenta circles) the South Evia main events (red circles). Beach balls show focal mechanisms of events (black color indicates compressional quadrants)

Full size image

Gomberg and Johnson (2005) had shown that after an earthquake not only numerous smaller shocks are triggered, over distances comparable to the dimensions of the mainshock fault rupture, but it can cause further earthquakes at any distance, if their amplitude exceeds several microstrain, regardless of their frequency content. Their results were confirmed by laboratory measurements of modulus reduction against dynamic loading strain amplitude for various rock types, pressures, and saturations.

Initially, in order to detect any indication of possible seismicity rate changes, we performed two statistical tests. The β-statistic (Matthews and Reasenberg 1988) is a measure of the difference between the number of events that occur versus the expected number of events normalized by the standard deviation:

$$\beta =\frac{{\textrm{N}}_{\textrm{a}}-\Lambda}{\sqrt{\Lambda}}$$

where \(\varLambda ={N}_b\times \frac{{\textrm{t}}_{\textrm{a}}}{{\textrm{t}}_{\textrm{b}}}\), Nb, and Na are the number of earthquakes before and after the stressing event (29 November 2022 events), respectively, and tb and ta are the time window lengths before and after, respectively. A β≥ 2.0 indicates a change in seismicity with 95% significance. The Z-statistic (Habermann 1987) defines the difference between the mean rate of seismicity in the period before and after the stressing event as

$$\textrm{Z}=\frac{\frac{{\textrm{N}}_{\textrm{a}}}{{\Delta \textrm{t}}_{\textrm{a}}}-\frac{{\textrm{N}}_{\textrm{b}}}{{\Delta \textrm{t}}_{\textrm{b}}}}{\sqrt{\frac{{\textrm{N}}_{\textrm{a}}}{{\Delta \textrm{t}}_{\textrm{a}}^2}+\frac{{\textrm{N}}_{\textrm{b}}}{{\Delta \textrm{t}}_{\textrm{b}}^2}}}$$

Z ~2 indicates 95% significance; thus, to indicate a change in the seismicity rate, we use a threshold of Z ≥ 2.0 (Pena Castro et al. 2019).

We considered tb = ta set to 60 days and we investigated the sequences, which took place afterwards in central Evia, independently (Fig. 10). The results are βΑ=30 and βB=62.7 while ZA = 7.3 and ZB = 12.4. Since β>2.0 and Z>2.0 for both activated areas, A and B, there is a certain indication of change in the seismicity rate which occurred during the November–December 2022 period (Fig. S13). Moreover, significant triggered rate changes implying dynamic triggering, has been observed in regions with geothermal activity (Gomberg et al. 2003; Brodsky 2006), a characteristic that may be found also in this case as central-north Evia is characterized by a shallow geothermal system in interaction with the local fault systems, both onshore and offshore (Palyvos et al. 2006; Karastathis et al. 2011; Kanellopoulos et al. 2016, 2020; Caroir et al. 2023). Consequently, we investigate the possibility of delayed dynamic stress triggering of the central Evia December 2022 events (areas A and B; Fig. 10) by the South Evia November 2022 major events.

When earthquake rupture propagates in a preferential direction (Haskell 1964), the ground motions in that direction can be greatly amplified and such a rupture directivity effect is frequently observed in not only large but also many moderate. In the case of L’Aquila2009 earthquake, Convertito et al. (2013) observed that for each main earthquake its aftershocks and the subsequent main events occur in the areas oriented as the source dominant rupture direction.

We measured the horizontal component of the seismic waves’ peak ground-motion velocity (PGV; Fig. 11) recorded at distances up to 280 km from the 29 November 2022 events (with magnitude, Mw, of 4.6 to 4.8); at each station, the PGV corresponds to the largest value between the two horizontal components of the recorded velocity. Figures 11 (a and b) show the map distribution of the PGV values for both events; a pattern of PGV values distribution can be identified, while largest values appear in a direction NNW-SSE, parallel to the fault rupture (see yellow line in Fig. 9). In the region of central Evia (Fig. 11), for both events, the PGVs approximate 0.4 cm/s that is expected to dynamically trigger faults at distances 60–80 km (Gomberg and Johnson 2005). Previous studies (Gomberget al. 2003) also have noted the influence of the rupture direction in the dynamic stress field for large and small-to-moderate earthquakes, expecting the greatest correlation for strike–slip events because the rupture propagation direction corresponds to maxima in the shear wave radiation pattern. In the near field, S waves should carry the largest dynamic strains (Boatwright et al. 2001) and PGV approximates the peak shear strain when divided by the shear wave velocity (Kanamori and Brodsky 2004). Peak dynamic strain is roughly proportional to the amplitude of seismic waves

$$\varepsilon \approx \frac{A}{\varLambda}\approx \frac{V}{C_s}$$

where A is displacement amplitude, Λ is wavelength, V is particle velocity, and CS is seismic wave velocity (Love 1927). Assuming radial symmetry (i.e., approximating the trigger as a point source), the peak dynamic strain in the near-field disk could be estimated as:

$$\varepsilon =\frac{PGV}{C_s}$$

where Cs is shear wave velocity and PGV is the peak ground velocity (van der Elst and Brodsky 2010). The velocity Cs refers to the depth of the hypocenter of the main source. This direct comparison of strain with velocity can highlight physical path effects (Farghal et al. 2020). The distribution of the peak PGV values (Fig. 11) indicates that the December 2022 seismicity rate in central Evia (areas A and B in Fig. 10) could be affected by the South Evia sequence. Peak dynamic strains range between 8 and 14 μstrain (Fig. 12).

Fig. 11
figure 11

Maps showing the peak ground velocity (PGV) distribution for the event of 29 November 2022 04:32 UTC (a) and for the event of 29 November 2022 20:06 UTC (b), presented as red stars. The December 2022 events that were possibly triggered afterwards in central Evia appear with blue stars

Full size image
Fig. 12
figure 12

Maps showing the peak dynamic strain (PDS) field obtained using PGV as a proxy calculated for a event of 29 November 2022 04:32 UTC and b event of 29 November 2022 20:06 UTC (red stars). The December 2022 events that were possibly triggered in central Evia appear with blue stars

Full size image

Moreover, we calculated the measured peak ground-motion velocity (PGV, which approximates the peak shear strain when divided by the shear wave or phase velocity) at distances normalized by rupture dimensions, using the square root of the rupture area (same area as in the Coulomb stress change calculations in section 4.1) as suggested by Gomberg and Johnson (2005). Evidence presented by Gomberg et al. (2001) imply that transient dynamic deformations could trigger earthquakes, although the mechanism(s) by which they do so remain unknown, while the thresholds span a small range from between 0.1cm/s to a few tens of cm/s and around 30microstrain (μs). Our observations (Fig. 13) indicate that dynamic triggering of the central Evia earthquakes might be possible.

Fig. 13
figure 13

Plot of measured peak ground velocity (PGVs, blue triangles; data from the NOA EIDA node) and strain (green triangles) against distance normalized by rupture dimensions for two main events of the sequence; a the event of 29 November 2022, 04:32 and b the event of 29 November 2022, 20:06

Full size image

In many cases triggering of earthquakes occurs within minutes to hours following the passage of the seismic waves (Brodsky et al. 2000; Fan et al. 2021), yet in other cases, earthquakes occurring weeks to months after the initial earthquake have been interpreted as delayed response to dynamic triggering (Hough 2005; Brodsky 2006; Parsons et al. 2014). Delayed triggered responses may reflect more complex series of physical processes and may trigger an aseismic process such as fault creep (Prejean and Hill 2009). Moreover, De Barros et al. (2017) found that regional earthquakes originating from azimuths between N340° and N80° with respect to the E-W trending Gulf of Corinth rift lead to dynamic triggering, whatever the distance or the wave amplitude. De Barros et al. (2017) suggested that earthquakes of smaller magnitudes also have the potential to trigger remote seismicity, while more delayed triggering might involve fluid diffusion (even the very small perturbations from remote Mw=4.5 earthquakes allow for a pressurization of interstitial fluids and an increase of seismicity). The December 2022–January 2023 (triggered) seismicity on central Evia occurred on an E-W striking normal fault zone (Psachna; Benetatos et al. 2004) and on N51°E (NOA solution) striking right-lateral strike-slip fault (Vlahia), respectively. We note that these faults are located to the NW of the November 2022 source events or in terms of azimuth they are oriented N300°E–N320°E with respect to the location of the triggering source. Moreover, the azimuthal location of the triggered events is nearly co-planar to the NW-SE orientation of the 29 November 2022 left-lateral, strike-slip fault, implying a possible directivity effect. Therefore, the delayed triggering (~30 days) of the central Evia events is not unexpected and may be explained by secondary triggering mechanisms initiated by dynamic stresses, such as fluid migration (De Barros et al. 2017). The upper crust of central and North Evia hosts geothermal fluids (e.g., Karastathis et al. 2011; Kanellopoulos et al. 2016, 2020) and transient, dynamic strains (Fig. 12) can induce transient changes in pressure. However, we note that quantifying the connection between earthquakes and delayed triggered seismic sequences at distances more than 2–3 rupture lengths is much more challenging.

6 Conclusions

We investigated the seismicity patterns preceding, during and following the 29 November 2022 events onshore South Evia, central Greece. The main findings can be summarized as follows:

  • The 29 November 2022 events ruptured a blind, near-vertical left-lateral strike slip fault striking NW-SE.

  • The fault plane is projected onshore South Evia southeast of village Almyropotamos.

  • The main events were preceded by a 35-day foreshock sequence that occurred in two groups with most events concentrated within 1-km from the epicenter of the 29 November 2022 04:32 UTC.

  • The frequency-magnitude distribution of the foreshocks is notably different than the post-mainshock sequence as the foreshock b-value = 1.37 vs. 0.78 of the latter.

  • The static stress transfer modeling indicated that stress loading on nearby faults diminishes below 0.1 bar at distances larger than 7 km from the mainshock.

  • The cumulative loading of the area SE of the epicenter of Mw=4.8 event (29 November 2022 20:06 UTC) reached stress levels which explain the occurrence of aftershocks for several months (end of April 2023).

  • The map distribution of computed PGVs and peak dynamic strains indicates local maxima in central Evia (Psachna and Vlahia regions) where two seismic sequences occurred in late December 2022–early January 2023.

  • The December 2022 sequences in central Evia possibly occurred because of delayed dynamic stress triggering due to the existence of geothermal fluids and the NW-SE strike (directivity effect) of the November 2022 seismic fault.