SESolid EarthSESolid Earth1869-9529Copernicus GmbHGöttingen, Germany10.5194/se-6-173-2015High-precision relocation of seismic sequences above a dipping Moho:
the case of the January–February 2014 seismic sequence on Cephalonia island
(Greece)KarastathisV. K.karastathis@noa.grMouzakiotisE.GanasA.https://orcid.org/0000-0002-1937-3283PapadopoulosG. A.https://orcid.org/0000-0003-1982-8214National Observatory of Athens, Institute of Geodynamics,
Lofos Nymfon, P.O. Box 20048, 11810 Athens, GreeceV. K. Karastathis (karastathis@noa.gr)12February2015611731845August20142September201415December201423December2014This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://www.solid-earth.net/6/173/2015/se-6-173-2015.htmlThe full text article is available as a PDF file from https://www.solid-earth.net/6/173/2015/se-6-173-2015.pdf
Detailed velocity structure and Moho mapping is of crucial importance for a
high precision relocation of seismicity occurring out of, or marginal to,
the geometry of seismological networks. Usually the seismographic networks do
not cover the boundaries of converging plates such as the Hellenic arc. The
crustal thinning from the plate boundary towards the back-arc area creates
significant errors in accurately locating the earthquake, especially when
distant seismic phases are included in the analysis. The case of the
Cephalonia (Ionian Sea, Greece) sequence of January–February 2014 provided
an excellent example where the hypocentral precision was greatly affected by
the crustal thinning from the plate boundary at the Ionian sea towards the
Aegean sea. This effect was examined in detail by testing various velocity
models of the region in order to determine an optimal model. Our tests
resulted in the adoption of a velocity model that resembles the crustal
thinning of the region. Then, a relocation procedure was performed in the
Cephalonia sequence for the time period of 26 January to 15 May 2014 by
applying probabilistic non-linear location algorithms. The high-precision
relocation resulted in an improved spatial distribution of the seismicity
with respect to the preliminary locations and provided a reliable basis to
examine seismotectonic implications of the Cephalonia sequence.
Introduction
On 26 January (13:55:42 and 18:45:08 GMT) and 3 February 2014
(03:08:44 GMT), western Cephalonia, an island on the Ionian Sea (Greece) (Fig. 1), was
ruptured by three strong earthquakes of magnitudes Mw=6.0,
Mw=5.3, and Mw=5.9 (Table 1, Fig. 2).
The two strongest earthquakes caused considerable damage in buildings and
infrastructure as well as several types of ground failures (rockfalls,
landslides, soil liquefaction) on the Paliki peninsula, mainly in Lixouri
and its surrounding villages (Papadopoulos et al., 2014; Valkaniotis et
al., 2014) (Fig. 1). The peak ground acceleration (PGA) recorded in several
localities at accelerometric stations, operated by the National Observatory
of Athens, Institute of Geodynamics (NOAGI) (NOA web report, 2014a, b) and
the Institute of Engineering Seismology and Earthquake Engineering
(EPPO-ITSAK, 2014a, b) reached up to 0.56 and 0.77 g in Lixouri during the
first and third earthquake, respectively. Since only three permanent seismic
stations existed in Cephalonia, on 28 and 29 January 2014 four
portable seismic stations (Fig. 2) were installed by NOAGI in the aftershock
zone of western Cephalonia with the aim of improving the monitoring
capabilities.
Location map of the study region. The Cephalonia Transform Fault
zone (CTFZ) is also indicated on the map. The seismographic stations of the
Hellenic Unified Seismological Network (HUSN) are depicted with green-colored
triangles and the strong motion stations with blue ones. Significant
earthquakes after 1900 (Papazachos et al., 2010) are shown as white circles:
(1) 24 January 1912, M=6.8; (2) 9 August 1953, M=6.4; (3) 11 August
1953, M=6.8; (4) 12 August 1953, M=6.3; (5) 12 August 1953, M=7.2;
(6) 17 September 1972, M=6.3; (7) 17 January 1983, M=7.0.
Cephalonia has been hit by many strong earthquakes in the past (Ambraseys,
2009; Papazachos and Papazachou, 2003). In the instrumental era of
seismology, the most important activity was a series of very strong, lethal
earthquakes that ruptured eastern and central Cephalonia, with the largest
(Ms=7.2) occurring on 12 August 1953. Large earthquakes occurred
off the coast of western Cephalonia in 1972 and 1983 (Scordilis et
al., 1985). The very high seismicity of Cephalonia is controlled by two major
seismotectonic structures. The first is the right-lateral strike-slip
Cephalonia Transform Fault zone (CTFZ) comprising of the NNE–SSW trending
Lefkada segment in the north and the NE–SW trending Cephalonia segment in
the south (Louvari et al., 1999) (Fig. 1). The strong (Mw=6.2)
earthquake on 14 August 2003 ruptured off the coast of western Lefkada along
the Lefkada segment of CTFZ (Fig. 1; Papadopoulos et al., 2003). Recently,
Papadopoulos et al. (2014) proposed, based on the spatial pattern of the
Cephalonia 2014 earthquake sequence, that the Lefkada segment does not
terminate at the NW side of Cephalonia, as previously thought, but extends
into western Cephalonia (Fig. 1). A second major structure that controls the
seismicity of the area is the northeastern subduction of the Ionian segment
of the Hellenic Arc beneath Cephalonia (Sachpazi et al., 2000), thus making
up a highly complex seismotectonic setting.
The January–February 2014 seismic sequence is the first one that ruptured
western Cephalonia and was recorded by modern seismograph instruments.
Therefore, the study of this sequence is of particular importance to better
understand the seismicity of Cephalonia. To this aim we performed a
high-precision relocation of the seismic sequence of more than 3300 events of
magnitude range M=1.0–6.0 from 26 January to 15 May 2014. Then,
relocation results were utilized to interpret the seismotectonics of the 2014
activity as well as the geometry and kinematics of the major structure of the
CTFZ.
Focal parameters of the three strong earthquakes of the
January–February 2014 Cephalonia seismic sequence as preliminary calculated
by NOA. The revised solutions are also shown.
Map of the aftershock sequence from 26 January to 15 May 2014 as
determined by the National Observatory of Athens. The moment tensor solutions
for the largest events as calculated by the Global Centroid Moment Tensor Project: Harvard University, USA
(HARV), National Observatory of Athens (NOA), and the German Research Centre
for Geoscience (GFZ).
The problem of location
The routine determination of earthquake hypocentral parameters usually
suffers from significant errors.
More specifically, the main sources of errors for an accurate determination of
the hypocentral parameters are (a) picking errors, (b) false
identification of the seismic phases, (c) insufficient number of phases,
(d) deficient azimuthal coverage of the seismographic network, and finally
(e) use of non-effective seismic velocity models that are usually
oversimplified (often 1-D) without adequate information for the
velocity structure and the lateral velocity heterogeneities. It is a common
practice that unified, regional 1-D velocity models are in use, which is also
the case of NOAGI for the daily seismicity monitoring of Greece (see Fig. 3).
Such velocity models deviate considerably from the local velocity structure
of an area, especially outside of the area covered by the network.
Although it is feasible to derive reliable seismic velocity models for areas
on a local scale through the implementation of nonlinear inversion
techniques that simultaneously invert travel-time data for the hypocentral
parameters and seismic velocity determination (Kissling et al., 1994;
Kissling, 1995; Thurber, 1993; Koulakov, 2009), it is difficult to perform
this practice on a regional scale. Additionally, information coming from
crustal seismic surveys is usually useful to build only local velocity
models. The crustal mapping based on gravity models suffers from intrinsic
ambiguity and the resolution of the seismic velocity global models is too low
(Koulakov and Sobolev, 2006).
Inaccurate hypocenter determinations become more acute in the presence of
strong structural anomalies and variations that make a structure
different from the horizontally layered Earth. For example, crustal
thickness is strongly affected in areas situated in the vicinity of
convergent plate boundaries. This is the case for the thick continental Aegean
crust in the vicinity of the Hellenic subduction zone. In fact, the
compressional regime along the Hellenic Arc led to folding and thin-skinned
tectonics as well as to the creation of the Mediterranean ridge, which
evolved to an accretionary prism, and to subsequent thickening of the crust
(Underhill, 1989; Yem et al., 2011). Crustal surveys have shown that in
western Greece, where the oceanic crust of the African plate is sliding
beneath the Aegean area, the continental crustal thickness exceeds 40 km and
becomes progressively thinner to the east (Makris, 1978; Tsokas and Hansen,
1997; Papazachos and Nolet, 1997; Tiberi et al., 2001; Karagianni et
al., 2005; Pearce et al., 2012; Makris et al., 2013). At the southern Aegean Sea
the crustal thickness reaches values as low as 20 km or less (Makris,
1975, 1976, 1977; Bohnhoff et al., 2001; Tirel et al., 2004). A similar
pattern of crustal thinning to the east but with shallower Moho depths (28 km
under Cephalonia) was proposed by Sodoudi et al. (2006). Thus, when an
earthquake occurs in the thick part of the crust and the paths of the
first arriving waves pass through the Moho as it progressively becomes
shallower, the travel-time errors may increase considerably with the increase
of the epicentral distance. In contrast, shallower events are not so strongly
affected, particularly in short epicentral distances since only Pg phases are
actually picked. This structure causes an asymmetrical shape to the head-wave
wavefront. Due to this structure, the adoption of a 1-D velocity model (see
NOA-IG model in Fig. 3) can cause systematic travel-time residuals at the
event's location.
The seismic P- and S-wave velocity models tested for the
relocation of the aftershock sequence. Green, blue, and red correspond to
velocity models proposed by Haslinger et al. (1999), Sachpazi et al. (2000),
and Hirn et al. (1996), respectively. The model routinely used by NOAGI is
marked in purple. The S-wave model of NOAGI is based on a constant
Vp/Vs=1.73. For Hirn et al. (1996) and Sachpazi et
al. (2000), a constant Vp/Vs ratio of 1.80 was used as
proposed by previous studies (Hatzfeld et al., 1995; Le Meur et al., 1997;
see also Haslinger et al., 1999). Haslinger et al. (1999) provide a separate
velocity model.
Analysis of seismological data
The Cephalonia 2014 seismic sequence was examined exactly in this context.
The first strong earthquake on 26 January 2014 was recorded by the permanent
stations of the HUSN (Hellenic Unified Seismological Network, 2015) before
the installation of the portable network in Cephalonia. The focus of that
event was preliminary determined by NOAGI at a location situated about 5 km
NNE from the city of Argostoli and at a focal depth of h=21km
(Fig. 2) (Table 1). This location clearly falls outside of the aftershock
cloud as well as far from Lixouri, where the macroseismic epicenter was
determined according to the field observations performed by Papadopoulos et
al. (2014) (Fig. 2). The epicentral area is situated at the geometrical edge
of the permanent network, meaning that no stations were situated to
the west of the earthquake focus. Consequently, the azimuthal coverage was
poor, leading to unstable location solutions. In addition, the preliminary
solution included several phases from significantly large epicentral
distances extending up to 360 km, where the crustal thickness was
significantly smaller than that in the subduction regime of western
Cephalonia. The 1-D model used by NOAGI (Fig. 3), with the Moho boundary
placed at 40 km, could not match Pn phase data from areas of thin
crust, particularly at large epicentral distances. Therefore, significant
errors were involved and consequently the epicenter of the first strong
earthquake was shifted substantially to the east. Its aftershocks, however,
were not shifted significantly because their small magnitudes prevented the
use of many distant stations where the velocity model was inadequate. In a
next paragraph it is explained more precisely how errors are introduced in
the epicentral solution.
The preliminary epicenter of the strong (Mw=5.3)
aftershock on 26 January 2014 (18:45:08 GMT) was
located in a noticeably better agreement to the macroseismic epicenter
(Fig. 2),
although it was calculated before the installation of the portable network in
Cephalonia. This is explained by the fact that its actual focal depth was
shallower than that of the first strong earthquake. It is also worth
mentioning that 1 s after the first strong shock on 26
January 2014, another strong aftershock occurred. Unfortunately, the waveform
of this event partially overlapped with the waveform of the first strong
shock, making an accurate estimation of its local and moment magnitudes quite
difficult. The estimation of its duration magnitude is Md=5.0. The
existence of this event, although not reported in earthquake catalogs,
justifies the accounts of local people of another shock strongly felt in Cephalonia a few seconds after the
first major event.
After the installation of the portable network in western Cephalonia, the
maximum azimuthal gap was decreased significantly, becoming less than
180∘ for most of the seismic events. The preliminary focus of the strong
earthquake (Mw=5.9) on 3 February 2014 was located at a shallower depth
(h=11km, Table 1) but within the aftershock cloud (Fig. 2) and at a
position close to its macroseismic epicenter, which again was determined to
be
(Papadopoulos et al., 2014) in Lixouri.
To effectively relocate the hypocenter data of the sequence one may use a
velocity model resembling as much as possible the real velocity structure. An
effective model can compensate for most of the systematic time residuals
created at distant stations. An alternative method is to use a 1-D model
to find the epicentral distance range where the effect of the
inclined Moho does not considerably affect the location accuracy. This
approach limits the seismic phases taken into account. It is reasonably
expected that the two different approaches should lead to quite similar
results.
Selection of a velocity model – relocation of the Cephalonia 2014 earthquake sequenceProposed models for the region
For western Greece, including Cephalonia, few seismic velocity models have
been proposed (Hirn et al., 1996; Haslinger et al., 1999; Sachpazi et
al., 2000). The model proposed by Hirn et al. (1996) was directly based on
the results of the crustal seismic surveys carried out in 1992 within the
project STREAMERS. The profile ION-7, with an azimuth of N62, was conducted
offshore between Cephalonia and Zakynthos (Zante) islands, starting from the
deep Ionian basin and reaching the western Gulf of Patras for a total length
of 180 km (see maps in Hirn et al., 1996). For the data acquisition,
the motor vessel (M/V) Bin Hai 511 from Geco-Prakla was used with a
36-airgun array (for processing details see also in Kokinou et al., 2005).
The seismic reflection profile acquired provided useful information for the
shallower structure. However, no precise information was gathered for the
Moho interface. To get a rough estimate of the Moho depth, Hirn et al. (1996)
performed ray-tracing modeling of the wide-angle travel-time data recorded at
distant onshore stations positioned on the Greek mainland. Those stations
were located only at the eastern side of the profile. Further assumptions
were made for the velocity values beneath the 7 km depth. Thus, these
authors discussed a model with the lower crustal interface (Vp
between 5.8 and 6.8 kms-1) at 15 km depth and the Moho boundary
at 25 km. Since the structure in this model was almost horizontal,
the 1-D model in Fig. 3 can be easily derived.
The velocity model of Haslinger et al. (1999) (Fig. 3) was built for the
eastern region of Lefkada (western Greece), which as regards to the
Cephalonia 2014 sequence, concentrates a high percentage of the ray paths
between the earthquakes and the stations. This model was based on a passive
experiment and was built as a “1-D minimum velocity model” for this region
by the VELEST algorithm (Kissling et al., 1994; Kissling, 1995) and used in a
following stage as the initial model in the local earthquake tomography method
and SIMULPS code (Thurber, 1993; Eberhart-Phillips, 1990, 1993), implemented
to calculate the 3-D crustal velocity structure. The SIMULPS code uses a
linearized damped least-square inversion to solve the non-linear problem of
the hypocentral location and velocity model. Because of this non-linear
nature of the problem, the initial velocity model and the initial hypocenter
locations in the inversion procedure should be as close as possible to their
true values. The “1-D minimum velocity model” calculated by the VELEST
algorithm can provide a good approximation and be used as an initial
velocity model. The minimum 1-D velocity models are usually used for
seismicity relocation (e.g., Lippitsch et al., 2005; Ganas et al., 2014).
The model proposed by Sachpazi et al. (2000) (Fig. 3) was also created by the
VELEST algorithm in order to be used as the initial model in a 3-D local
earthquake tomography to determine the velocity structure of the studied
area. The procedure for the construction of a 1-D minimum velocity
model is highly dependent on the selection of an initial model (Karastathis
et al., 2011) and, therefore, it is usually based on the results of seismic
profiles. Sachpazi et al. (2000) based their initial model on the seismic
profiles presented by Hirn et al. (1996).
For the adoption of an appropriate seismic velocity model, we compared the
three 1-D models mentioned above (Fig. 3). As we will see later in detail, the
model that performed better was that proposed by Haslinger et al. (1999).
With a vertical velocity gradient based on Haslinger et al. (1999) we
constructed, tested, and adopted a 2-D velocity model with a non-horizontal
Moho boundary based on Papazachos and Nolet (1997). Figure 4 shows the vertical
cross section of the 2-D model.
The 2-D velocity model tested to assess the influence of Moho
structure. The Moho boundary has been based on the results of Papazachos and
Nolet (1997). The position of Cephalonia is between 0 and 50 km. The
hypocenters of the major events are shown with stars: blue for the event of
26 January 2014 (Mw=6.0), yellow for the aftershock on
26 January 2014 (Mw=5.3), and red for the event of 3 February 2014
(Mw=5.9).
Synthetic data
Before comparing the performance of these models, we examine the influence on
the location procedure of the non-horizontal Moho boundary in the Aegean
region. More specifically, in order to assess the impact of the errors
imposed in the earthquake location procedure by the adoption of a simplified
1-D model in the presence of a non-horizontal Moho structure, we constructed
synthetic arrival times for the adopted model using the 3-D version of the
eikonal finite-difference scheme of Podvin and Lecomte (1991) and estimated
the time differences for both a horizontal and a non-horizontal Moho
structure. The velocity gradient was based on Haslinger et al. (1999). The
2-D model with the non-horizontal Moho boundary is shown in Fig. 4 and the
results of the comparison with the respective 1-D model are depicted in
Fig. 5. As one may expect, the time difference is zero only when the first
arrivals are also zero due to the Pg seismic phases. Obviously, the shallower
events, with focal depth between 5 and 10 km are not, or only
slightly, affected, particularly when they do not bare enough energy to
travel at long epicentral distances. As a result, the majority of the shallow
aftershocks remain unaffected. In contrast, the influence is higher for the
deeper and stronger events, such as the first strong earthquake on 26 January
2014. For this strong earthquake three different epicenters were calculated
with the use of the 1-D (with the inclusion or the exclusion of distant
phases) and 2-D (with all phases) models (Fig. 6). It can be seen how the
simplified 1-D velocity model affects the epicenter location when distant
phases are taken into account. The error decreases when distant phases are
omitted.
Time differences of P-wave (the two panels in the upper row) and
S-wave (the two panels in the lower panel) arrivals between synthetic data
calculated on the basis of the 1-D model and data collected using the 2-D
model (based on the same 1-D model but with a non-horizontal Moho boundary).
Earthquake focal depths of 5 km (left column) and 15 km (right column) are
represented. The hypothetical epicenter is shown as a red circle. The errors
imported in the case which does not take into account the Moho structure can
be significant at long distances.
The epicenter of the strong earthquake (Mw=6.0) on
26 January 2014 calculated with the 1-D model of Haslinger et al. (1999) (see
Fig. 3) including distant phases (light blue star at the east), the same 1-D
model excluding distant phases (blue star at the west), and the 2-D model
(see Fig. 4) including distant phases (red star).
We concluded that, for lack of reliable knowledge regarding the structure and
velocity of the Moho boundary and in the presence of poor azimuthal
seismographic coverage, it is preferable to limit the range of the epicentral
distances of stations used and to base the location mostly on the Pg phases.
The data processing has been performed by the NonLinLoc algorithm (Lomax et
al., 2000) that follows a non-linear earthquake location method, giving a
complete probabilistic solution expressed in terms of a posterior density
function (PDF) (Tarantola and Valette, 1982). The function is calculated
using the equal differential time (EDT) likelihood function and depicted by
confidence ellipsoids. Therefore the higher the confidence of the velocity
model, the smaller the ellipsoids of the event locations. This
probabilistic approach is characterized by strong advantages
compared to linearized methods. More precisely, the EDT function provides a
more reliable uncertainty estimate, especially in the presence of outliers,
than the conventional least-squares L1 and L2 norms for the misfit
calculation between the observed and calculated travel times. Another
advantage of the method is that it is independent of origin time, so the 4-D
problem of hypocenter location reduces to a 3-D search over spatial location
of the hypocenter (latitude, longitude, depth). The NonLinLoc algorithm can
also use 2-D and 3-D velocity models.
Real data
For the relocation of the Cephalonia 2014 aftershock sequence, we used NOAGI
phase data set consisting of more than 44 000 P-wave and 24 000 S-wave
arrivals for the time interval from 26 January to 15 May 2014,
corresponding to more than 3300 events. Phase data from distant stations
(Δ>120km) were excluded. It should be noted, however,
that small magnitude events remained unaffected since they could hardly be
identified at greater epicentral distances anyway.
Relocation results by using the velocity models of (a) the
2-D velocity model that resembles the Moho structure, (b) the
1-D velocity model proposed by Haslinger et al. (1999),
(c) the 1-D velocity model proposed by Sachpazi et al. (2000),
(d) the 1-D velocity model proposed by Hirn et al. (1996), and
(e) the 1-D velocity model used by NOAGI for the daily seismic
monitoring.
We compared both the 2-D and 1-D versions of the adopted model with the three
models proposed by previous authors (Fig. 3). For the comparison we selected
events with at least six P- and one S-wave arrivals and an azimuthal gap
lower than 180∘. The station delays were calculated as an average
value of observed travel-time residual of the well-located events (GAP <180∘, RMS <1s, horizontal uncertainty ERH and vertical
uncertainty ERZ <1km) and applied to the location procedure.
Station corrections compensate for the effect of the station's local geology,
which could not be taken into account by the use of a 1-D velocity model. The
adopted model succeeds in producing a more compact horizontal projection
(Fig. 7) and verifies that the aftershock sequence, trending NNE–SSW, covers
only the western part of Cephalonia at a length of about 35 km and
maximum lateral width of about 10 km (see also Table 1 for the
relocated hypocenters of the three major events). It is noteworthy that the
relocated aftershock area nearly coincides with the main part of the
macroseismic field that is covered by the isoseismal of level V,
which is also the area of ground failures produced by the strong earthquakes
on 26 January and 3 February 2014 (Papadopoulos et al., 2014).
The model of Haslinger et al. (1999) performs significantly better than those
of Hirn et al. (1996), Sachpazi et al. (2000), and that of NOAGI because it
is derived from histograms of horizontal and vertical location uncertainties
(Table 2). The vast majority (about 80 %) of the events relocated with
the adopted model have a horizontal uncertainty less than 900 m.
Moreover, the 52 % of the relocated events have a horizontal uncertainty
less than 600 m (Table 2). As we can also see in the same table, the
original 1-D model of Haslinger et al. (1999) is not notably inferior to the
other three models, which produced significantly larger uncertainties.
Similar results can be seen also in Fig. 9, depicting the vertical
uncertainty distribution (see also Table 2).
Distribution of horizontal (ERH) and vertical (ERZ) uncertainties
for the events relocated with several seismic velocity models.
For the first 15 km of depth, the model of Haslinger et al. (1999)
has similar velocity values with the model proposed by Hirn et al. (1996),
which was produced from reliable data of seismic reflection profiles.
However, there is an obvious discrepancy at depths greater than
15 km. This might be explained by the fact that for these depths,
Hirn et al. (1996) used results only from ray-tracing modeling based on
common receiver data only from one side (eastern part) of the seismic
traverse. In contrast, Haslinger et al. (1999) did not consider an abrupt
increase in the velocity structure at 15 km depth and proposed a Moho
boundary at 40 km, whereas Hirn et al. (1996) used a Moho depth at
25 km. The shallow Moho boundary is likely the main reason for the
poor relocation results we obtained from the model of Hirn et al. (1996).
The aftershock focal depths calculated by various models also show
significant variations (Fig. 8). The adopted model, as well as that of
Haslinger et al. (1999), has the vast majority of focal depths between 6 and
14 km. The model of Sachpazi et al. (2000) calculated a significant
percentage of the hypocenters at depths between 4 and 6 km and a very
low percentage with depths greater than 12 km. No hypocenters with
unrealistic depth values (<4) were calculated by the models based on
Haslinger et al. (1999).
Seismotectonic implications and discussion
The spatial distribution of the relocated earthquake sequence (Fig. 7a)
confirms that the 2014 earthquake activity covers only the western part of
Cephalonia Island trending from NNE to SSW at a length of about 35 km
and maximum lateral width of ca. 10 km. No earthquake activity
developed off the coast of western Cephalonia. As a consequence, the
January–February 2014 earthquake sequence can hardly be seismotectonically
associated with the Cephalonia segment of the major right-lateral strike-slip
structure of the CTFZ, as the latter was proposed by Louvari et al. (1999).
On the contrary, the aftershock pattern implies that the 2014 activity
ruptured western Cephalonia due to onshore strike-slip faulting. One possible
scenario is that the activated strike-slip faults comprise the southern
prolongation of the NNE–SSW trending Lefkada segment of the CTFZ.
Papadopoulos et al. (2014) suggested that the Lefkada CTFZ segment does not
terminate off the coast of NW Cephalonia, as proposed by previous authors
(Louvari et al., 1999), but extends further into western Cephalonia. Another
scenario is that the activated strike-slip faults comprise segments of a
30 km, nearly N–S trending fault zone that splits the island into a
western and an eastern part. The consequence is that western Cephalonia
appears as a seismotectonic block independent of eastern Cephalonia, which
hosted the sources of the large 1953 earthquakes.
The space–time evolution of the 2014 sequence (Fig. 9), based on the
high-precision relocated earthquake catalogue that we produced, indicates
that soon after the occurrence of the first strong earthquake on 26 January
2014, the aftershock area was already well shaped. No further expansion of the
aftershock area was observed either after 29 January, when the
portable network was installed, or after the strong
earthquake on 3 February 2014.
Distribution of hypocentral depth of the relocated Cephalonia 2014
earthquake sequence up to 15 May 2014 for different seismic velocity models:
(a) the adopted 2-D model, which is a modification of the Haslinger
et al. (1999) model; (b) Haslinger et al. (1999);
(c) Sachpazi et al. (2000); (d) Hirn et al. (1996); and
(e) the model routinely used in the daily seismic monitoring by
NOAGI.
Space–time evolution of the Cephalonia 2014 sequence. The maps show
the aftershocks with 1-week time interval (a–f) between 26 January
2014 and 26 March 2014.
The Cephalonia 2014 sequence is geographically distributed into two clusters
(Figs. 7a and 10). The first is small, with a length on the order of
10 km and occupying the north side of the aftershock cloud. The other
extends in the central and south sides, thus leaving an apparent spatial gap
between the two clusters. Papadopoulos et al. (2014) suggested that the area
of the 2014 gap had already ruptured from the strong (Mw=5.5)
strike-slip earthquake on 25 March 2007. However, no temporal relation was
found between these two clusters and the occurrence of the strong events on
26 January and 3 February 2014. The north cluster abuts but does not overlap
the southern side of the aftershock area of the strong 2003 Lefkada
(Mw=6.2) main shock (Papadopoulos et al., 2003). Besides, the
foreshock activity that preceded the first strong earthquake on 26 January
2014 by about 4 days was recorded exactly in the area of the north (small)
aftershock cluster (Papadopoulos et al., 2014). This may indicate that the
2014 activity was initiated at the northern part of the aftershock area where
the 2003 Lefkada activity diminished. Therefore, we observe that a shallow
tectonic structure exists in the area of the Myrtos Gulf, possibly a
near-vertical fault striking WNW–ESE that is perpendicular to the NNE–SSW
strike of the Lefkada 2003 and the Cephalonia 2014 aftershock areas. This
fault, which probably controlled the initiation of the 2014 sequence, can be
seen in the NNE–SSW cross section in Fig. 10c that depicts clearly the
vertical geometry of the EW cross fault at Myrtos Gulf at depths of
5–12 km. Evidence also comes from the space–time evolution of the
relocated 2014 sequence (Fig. 9), but further examination is needed.
Mean horizontal and absolute source location distances between the
adopted model and the other seismic velocity models.
Haslinger et al. (1999)Sachpazi et al. (2000)Hirn et al. (1996)NOAGIMean horizontal distance (km)0.71.85.32.4Mean absolute distance (km)2.53.17.13.7
Vertical sections of the aftershock sequence and their location on
the map: (a) the hypocenters (with GAP <180∘) between 26
and 30 January 2014; (b) the hypocenters (with GAP <180∘)
between 3 and 8 February 2014; and (c) the hypocenters (with
GAP <180∘) for the whole sequence (26 January–15 May
2014).
With the relocation applied, the foci of the three strongest earthquakes of
the sequence shifted at shallower depths, while the first strong earthquake
on 26 January 2014 also shifted towards WNW with respect to the preliminary
determinations (Table 1). The thickness of the seismogenic layer does not
exceed 16 km. That the 2014 aftershock area was well formed from the
very beginning without spatial expansion after the strong earthquake on
3 February 2014 provides evidence that this earthquake ruptured within the
aftershock volume of the 26 January 2014 earthquake, the largest
event of the sequence. From this point of view we may consider that the
3 February event was the strongest aftershock of the sequence that ruptured
at a shallower depth and at a different fault from that of the main shock on
26 January 2014.
To further control the fault patterns associated with the 26 January and
3 February earthquakes, we constructed two respective vertical cross sections
as shown in Fig. 10. One may observe that in the section corresponding to the
main shock on 26 January, the aftershocks until 30 January included in a ±4 km wide zone seem arranged in a plane of nearly (+/-10∘) N–S
and dip of about 65∘ to the east (Fig. 10a). The preferred fault-plane
adopted by Papadopoulos et al. (2014) and Valkaniotis et al. (2014) is of
strike 23∘ and dip 68∘ to the east, which is consistent with
the geometry represented by the vertical section. However, the vertical
section through the hypocenter of the 3 February 2014 event (Fig. 10b) shows
that the fault plane strikes nearly (+/-10∘) N–S but its dip is
about 65∘ to the west. The geometry of this fault plane is compatible
with the fault plane that dips 56∘ to the west according to the focal
mechanism computed by the GFZ (183∘/56∘/138∘; reported on
the European-Mediterranean Seismological Centre website (2015). The western
dip in combination with the oblique-slip rake may result in uplift of the
hanging wall (western) of the N–S fault during co-seismic motion.
Our relocation procedure suggests a different fault model than that of
Karakostas et al. (2014; their Fig. 8), who suggested a right step of CTFZ,
on the Paliki peninsula. Contrarily, we detected the activation of two blind
strike-slip faults along the N–S axis at the Myrtos Gulf–Lixouri line,
possibly overlapping with a left step. The 26 January 2014 activated fault
(Fig. 10a) is in agreement with the blind fault model of Valkaniotis et
al. (2014). Such a fault configuration may explain the co-seismic uplift seen
on the Paliki peninsula in InSAR data (Boncori et al., 2015) because our
relocation data in the epicentral region of the 3 February 2014 event point
to a west-dipping fault (Fig. 10b). If this is the case, then the hanging
wall of this fault moved upwards during co-seismic motion; it is known that
Cephalonia region is in state of compression with maximum horizontal stress
oriented at N78∘ E (+/-9∘; Ganas et al., 2013), as
determined from GPS data. In addition, Lagios et al. (2012) obtained a N–S
discontinuity in their horizontal velocity field (see Fig. 4c of Lagios et
al., 2012) across the Gulf of Argostoli, which may be indicative of a crustal
block boundary or a large fault zone beneath the Gulf.
Conclusions
The strong earthquakes on Cephalonia on 26 January and 3 February 2014
clearly show the problem of incorrect hypocentral determination for
earthquakes that occur at the borders of the seismograph network of Greece,
near the region of convergence of the tectonic plates when an oversimplified
1-D velocity model is used.
We propose a 2-D velocity model with a non-horizontal Moho boundary that
resembles the crustal thinning of the Hellenic Arc region.
We applied a relocation procedure in the 2014 Cephalonia sequence for the
time period from 26 January to 15 May 2014 by applying probabilistic
non-linear location algorithms. The thickness of the seismogenic layer of the
upper (Aegean) plate does not exceed 16 km.
The distribution of the relocated epicenters covers only the western part
of Cephalonia, trending NNE–SSW for about 35 km with a lateral width of
ca. 10 km.
The 2014 earthquake sequence is associated with dextral, strike-slip faults
on the island of Cephalonia. Moreover, a vertical section through the hypocenter of
the 3 February 2014 event shows that the fault plane strikes nearly N–S but
its dip is about 65∘ to the west.
Our relocated hypocentral distribution delineates a shallow, near-vertical
fault existing in the area of Myrtos Gulf, striking WNW–ESE, perpendicular to
both the strike of the Lefkada 2003 and the Cephalonia 2014 aftershock areas.
Acknowledgements
Thanks are extended to the acting director and the staff of the Institute of
Geodynamics for the daily processing of many seismic phases as well as for
the installation of the portable network in Cephalonia. Thanks are also due
to the topical editor of the journal, Takaaki Taira, for his suggestions on
improving the original manuscript. We would also like thank Ivan Koulakov and
the anonymous reviewer for the thorough review of the manuscript and their
constructive comments.
Edited by: T. Taira
References
Ambraseys, N.: Earthquakes in the Mediterranean and Middle East, Cambridge
University Press, New York, 8, 4333–4355, 2009.
Bohnhoff, M., Makris, J., Papanikolaou, D., and Stavrakakis, G.: Crustal
investigation of the Hellenic subduction zone using wide aperture seismic
data, Tectonophysics, 343, 239–262, 2001.
Boncori, J. P. M., Papoutsis, I., Pezzo, G., Tolomei, C., Atzori, S., Ganas,
A., Karastathis, V., Salvi, S., Kontoes, C., and Antonioli, A.: The February
2014 Cephalonia Earthquake (Greece): 3D Deformation Field and Source Modeling
from Multiple SAR Techniques, Seismol. Res. Lett., in press, 2015.
Eberhart-Phillips, D.: Three-dimensional P and S velocity structure in the
Coalinga Region, California, J. Geophys. Res., 95, 15343–15363, 1990.
Eberhart-Phillips, D.: Local earthquake tomography: earthquake source
regions, in: Seismic Tomography: Theory and Practice, edited by: Iyer, H. M.
and Hiahara, K., Chapman and Hall, London, 613–643, 1993.EPPO-ITSAK: The January 26th, 2014 Earthquake in Cephalonia, Preliminary
Report, by EPPO-ITSAK, available at:
http://www.slideshare.net/itsak-eppo/2014-0127-kefaloniaeqpreliminaryreporta,
last access: 28 January 2015, 2014a (in Greek).EPPO-ITSAK: Strong Ground Motion of the February 3, 2014 Cephalonia
Earthquake: Effect on Soil and Built Environment in Combination with the
January 26, 2014 Event, available at:
http://www.slideshare.net/itsak-eppo/20140203-kefaloniaeq-report-en,
last access: 28 January 2015, 2014b.European-Mediterranean Seismological Centre website: Full tensors solutions
(event 20140203 03:08),
http://www.emsc-csem.org/Earthquake/mtfull.php?id=357329, last access:
28 January 2015.
Ganas, A., Marinou, A., Anastasiou, D., Paradissis, D., Papazissi, K.,
Tzavaras, P., and Drakatos, G.: GPS-derived estimates of crustal deformation
in the central and north Ionian Sea, Greece: 3-yr results from NOANET
continuous network data, J. Geodynamics, 67, 62–71, 2013.
Ganas, A., Karastathis, V., Moshou, A., Valkaniotis, S., Mouzakiotis, E., and
Papathanassiou, G.: Aftershock properties and seismotectonic setting of the 7
August 2013 moderate earthquake in Kallidromon Mountain, central Greece,
Tectonophysics, 617, 101–113, 2014.
Haslinger, F., Kissling, E., Ansorge, J., Hatzfeld, D., Papadimitriou, E.,
Karakostas, V., Makropoulos, K., Kahle, H. G., and Peter, Y.: 3D crustal
structure from local earthquake tomography around the Gulf of Arta (Ionian
region, NW Greece), Tectonophysics, 304, 201–218, 1999.
Hatzfeld, D., Kassaras, I., Panagiotopoulos, D., Amorese, D., Makropoulos,
K., Karakaisis G., and Coutant, O.: Microseismicity and strain pattern in
northwestern Greece, Tectonics, 14, 773–785, 1995.Hellenic Unified Seismological Network: Webpage of Hellenic Unified Seismic
Network (H.U.S.N.) hosted in the website of Institute of Geodynamics, of
National Observatory of Athens:
http://www.gein.noa.gr/en/networks/husn/, last access: 28 January 2015.
Hirn, A., Sachpazi, M., Siliqi, R., Bride, J. M., Marnelis, F., Cernobori,
L., and STREAMERS-PROFILES group: A traverse of the Ionian islands front with
coincident normal incidence and wide-angle seismics, Tectonophysics, 264,
35–49, 1996.
Karagianni, E. E., Papazachos, C. B., Panagiotopoulos, D. G., Suhadolc, P.,
Vuan, A., and Panza, G. F.: Shear velocity structure in the Aegean area
obtained by inversion of Rayleigh waves, Geophys. J. Int., 160, 127–143,
2005.
Karakostas, V., Papadimitriou, E., Mesimeri, M., Gkarlaouni, C., and
Paradisopoulou, P.: The 2014 Kefalonia doublet (Mw 6.1 and Mw 6.0), Central
Ionian Islands, Greece: Seismotectonic implications along the Kefalonia
transform fault zone, Acta Geophys., 1–16, 2014.
Karastathis, V. K., Papoulia, J., Di Fiore, B., Makris, J., Tsambas, A.,
Stampolidis, A., and Papadopoulos, G. A.: Deep structure investigations of
the geothermal field of the North Euboean Gulf, Greece, using 3-D local
earthquake tomography and Curie point depth analysis, J. Volcanol. Geoth.
Res., 206, 106–120, 2011.Kissling, E.: Velest User's Guide, Institute of Geophysics, ETH Zurich,
Switzerland, 26 pp., http://www.seg.ethz.ch/software/velest_guide.pdf,
1995.
Kissling, E., Ellsworth, W. L., Eberhart-Phillips, D., and Kradolfer, U.:
Initial reference models in local earthquake tomography, J. Geophys. Res.,
99, 19635–19646, 1994.
Kokinou, E., Kamberis, E., Vafidis, A., Monopolis, D., Ananiadis, G., and
Zelelidis, A.: Deep seismic reflection data Greece: a new crustal model for
the Ionian sea, J. Petrol. Geol., 28, 81–98, 2005.
Koulakov, I.: LOTOS code for local earthquake tomographic inversion:
benchmarks for testing tomographic algorithms, B. Seismol. Soc. Am., 99,
194–214, 2009.
Koulakov, I. and Sobolev, S. V.: Moho depth and three-dimensional P and S
structure of the crust and uppermost mantle in the Eastern Mediterranean and
Middle East derived from tomographic inversion of local ISC data, Geophys. J.
Int., 164, 218–235, 2006.
Lagios, E., Papadimitriou, P., Novali, F., Sakkas, V., Fumagalli, A.,
Vlachou, K., and Del Conte, S.: Combined Seismicity Pattern Analysis, DGPS and
PSInSAR studies in the broader area of Cephalonia (Greece), Tectonophysics,
524, 43–58, 2012.
Le Meur, H., Virieux, J., and Podvin, P.: Seismic tomography of the Gulf of
Corinth: a comparison of methods, Ann. Geophys.-Italy, 40, 1–24, 1997.
Lippitsch, R., White, R. S., and Soosalu, H.: Precise hypocentre relocation of
microearthquakes in a high temperature geothermal field: the Torfajökull
central volcano, Iceland. Geophys. J. Int, 160, 371–388, 2005.
Lomax, A., Virieux, J., Volant, P., and Berge-Thierry, C.: Probabilistic
earthquake location in 3-D and layered models, in: Advances in Seismic Event
Location, edited by: Thurber, C. H. and Rabinowitz, N., Kluwer Academic
Publishers, Dordrecht/Boston/London, 101–134, 2000.
Louvari, E., Kiratzi, A. A., and Papazachos, B. C.: The CTF and its extension
to western Lefkada Island, Tectonophysics, 308, 223–236, 1999.
Makris, J.: Crustal Structure of the Aegean Sea and the Hellenides Obtained
from Geophysical Survey, J. Geophys., 41, 441–443, 1975.
Makris, J.: A dynamic model of the Hellenic arc deduced from geophysical
data, Tectonophysics, 36, 339–346, 1976.
Makris, J.: Geophysical Investigations of the Hellenides, Hamburger
Geophysikalische Einzelschriften, G. M. L. Wittenborn Söhne, Hamburg,
Heft 34, 124 pp., 1977.
Makris, J.: The crust and upper mantle of the Aegean region from deep seismic
soundings, Tectonophysics, 46, 269–284, 1978.
Makris, J., Papoulia, J., and Yegorova, T.: A 3-D density model of Greece
constrained by gravity and seismic data, Geophys. J. Int., 194, 1–17, 2013.NOA web report, internal report, available at:
http://www.gein.noa.gr/Documents/pdf/Cefalonia_20140126_preliminary_web.pdf
(last access: 4 June 2014), 2014a.NOA web report, internal report, available at:
http://www.gein.noa.gr/Documents/pdf/Cefalonia_20140203_preliminary_web.pdf
(last access: 4 June 2014), 2014b.
Papadopoulos, G. A., Karastathis, V. K., Ganas, A., Pavlides, S., Fokaefs,
A., and Orfanogiannaki, K.: The Lefkada, Ionian Sea (Greece), shock (Mw 6.2)
of 14 August 2003: Evidence for the characteristic earthquake from seismicity
and ground failures, Earth Planet. Space, 55, 713–718, 2003.
Papadopoulos, G. A., Karastathis, V. K., Koukouvelas, I., Sachpazi, M.,
Baskoutas, I., Chouliaras, G., Agalos, A., Daskalaki, E., Minadakis, G.,
Moschou, A., Mouzakiotis, E., Orfanogiannaki, A., Papageorgiou, A., Spanos
D., and Triantafyllou, I.: The Cephalonia, Ionian Sea (Greece), sequence of
strong earthquakes of January-February 2014: a first report, Res. Geophys.,
4, 19–30, 2014.Papazachos, B. C. and Papazachou C. B.: The Earthquakes of Greece, Ziti
Editions, Thessaloniki, 2003 (in Greek).
Papazachos, B. C., Comninakis, P. E., Scordilis, E. M., Karakaisis, G. F.,
and Papazachos C. B.: A catalogue of earthquakes in the Mediterranean and
surrounding area for the period 1901–2010, Publ. Geophys. Laboratory,
University of Thessaloniki, Greece, 2010.
Papazachos, C. B. and Nolet, G.: P and S deep velocity structure of the
Hellenic area obtained by robust nonlinear inversion of travel times, J.
Geophys. Res., 102, 8349–8367, 1997.
Pearce, F. D., Rondenay, S., Sachpazi, M., Charalampakis, M., and Royden,
L. H.: Seismic investigation of the transition from continental to oceanic
subduction along the western Hellenic Subduction Zone, J. Geophys. Res., 117,
1978–2012, 2012.
Podvin, P. and Lecomte, I.: Finite difference computation of traveltimes in
very contrasted velocity models: a massively parallel approach and its
associated tools, Geophys. J. Int., 105, 271–284, 1991.
Sachpazi, M., Hirn, A., Clement, C., Laigle, M., Haslinger, F., Kissling,
E., Charvis, P., Hello, Y., Lepine, J., Sapine, M., and Ansorge, J.: Western
Hellenic subduction and Cephalonia transform: local earthquakes and plate
transport and strain, Tectonophysics, 319, 301–319, 2000.
Scordilis, E. M., Karakaisis, G. F., Karakostas, B. G., Panagiotopoulos,
D. G., Comninakis, P. E., and Papazachos, B. C.: Evidence for Transform
Faulting in the Ionian Sea: the Cephalonia Island Earthquake Sequence of
1983, Pure Appl. Geophys., 123, 388–397, 1985.Sodoudi, F., Kind, R., Hatzfeld, D., Priestley, K., Hanka, W., Wylegalla,
K., Stavrakakis, G., Vafidis, A., Harjes H.-P., and Bohnhoff, M.: Lithospheric
structure of the Aegean obtained from P and S receiver functions, J. Geophys.
Res., 111, B12307, 10.1029/2005JB003932, 2006.Tarantola, A. and Valette, B.: Inverse problems = quest for information,
J. Geophys., 50, 159–170, 1982.Thurber, C. H.: Local earthquake tomography: velocities and Vp=Vs – theory, in: Seismic Tomography: Theory and Practice, edited
by: Iyer, H. M. and Hiahara, K., Chapman and Hall, London, 563–583, 1993.
Tiberi, C., Diament, M., Lyon-Caen, H., and King, T.: Moho topography beneath
the Corinth Rift area (Greece) from inversion of gravity data, Geophys. J.
Int., 145, 797–808, 2001.
Tirel, C., Gueydan, F., Tiberi, C., and Brun, J. P.:. Aegean crustal
thickness inferred from gravity inversion. Geodynamical implications, Earth
Planet. Sc. Lett., 228, 267–280, 2004.
Tsokas, G. N. and Hansen, R. O.: Study of the crustal thickness and the
subducting lithosphere in Greece from gravity data, J. Geophys. Res., 102,
20585–20597, 1997.
Underhill, J. R.: Late Cenozoic deformation of the Hellenide foreland,
western Greece, Geol. Soc. Am. Bull., 101, 613–634, 1989.Valkaniotis, S., Ganas, A., Papathanassiou, G., and Papanikolaou, M.: Field
observations of geological effects triggered by the January-February 2014
Cephalonia (Ionian Sea, Greece) earthquakes, Tectonophysics, 630, 150–157,
10.1016/j.tecto.2014.05.012, 2014.
Yem, L. M., Camera, L., Mascle, J., and Ribodetti, A.: Seismic stratigraphy
and deformational styles of the offshore Cyrenaica (Libya) and bordering
Mediterranean Ridge, Geophys. J. Int., 185, 65–77, 2011.