Content uploaded by Aybige Akinci
Author content
All content in this area was uploaded by Aybige Akinci on Feb 10, 2023
Content may be subject to copyright.
Geophys. J. Int. (2020) 223, 1054–1068 doi: 10.1093/gji/ggaa350
Advance Access publication 2020 July 21
GJI Seismology
Source modelling and strong ground motion simulations for the 24
January 2020, Mw6.8 Elazı˘
g earthquake, Turkey
Daniele Cheloni and Aybige Akinci
Istituto Nazionale di Geofisica e Vulcanologia, Via di Vigna Murata 605,00143 Roma, Italy. E-mail: daniele.cheloni@ingv.it
Accepted 2020 July 18. Received 2020 June 23; in original form 2020 May 4
SUMMARY
On 24 January 2020 an Mw6.8 earthquake occurred at 20:55 local time (17:55 UTC) in eastern
Turkey, close to the town of Sivrice in the Elazı˘
g province, causing widespread considerable
seismic damage in buildings. In this study, we analyse the main features of the rupture process
and the seismic ground shaking during the Elazı˘
g earthquake. We first use Interferometric
Synthetic Aperture Radar (InSAR) interferograms (Sentinel-1 satellites) to constrain the fault
geometry and the coseismic slip distribution of the causative fault segment. Then, we utilize
this information to analyse the ground motion characteristics of the main shock in terms of
peak ground acceleration (PGA), peak ground velocity (PGV) and spectral accelerations. The
absence of seismic registrations in near-field for this earthquake imposes major constraints on
the computation of seismic ground motion estimations in the study area. To do this, we have
used a stochastic finite-fault simulation method to generate high-frequency ground motions
synthetics for the Mw6.8 Elazı˘
g 2020 earthquake. Finally, we evaluate the potential state of
stress of the unruptured portions of the causative fault segment as well as of adjacent segments,
using the Coulomb stress failure function variations. Modelling of geodetic data shows that
the 2020 Elazı˘
g earthquake ruptured two major slip patches (for a total length of about 40 km)
located along the P ¨
ut¨
urge segment of the well-known left-lateral strike-slip East Anatolian Fault
Zone (EAFZ), with up to 2.3 m of slip and an estimated geodetic moment of 1.70 ×1019
Nm (equivalent to a Mw6.8). The position of the hypocentre supports the evidence of marked
WSW rupture directivity during the main shock. In terms of ground motion characteristics, we
observe that the high-frequency stochastic ground motion simulations have a good capability to
reproduce the source complexity and capture the ground motion attenuation decay as a function
of distance, up to the 200 km. We also demonstrate that the design spectra corresponding to
475 yr return period, provided by the new Turkish building code is not exceeded by the
simulated seismograms in the epicentral area where there are no strong motion stations and no
recordings available. Finally, based on the Coulomb stress distribution computation, we find
that the Elazı˘
g main shock increased the stress level of the westernmost part of the P¨
ut¨
urge
fault and of the adjacent Palu segment and as a result of an off-fault lobe.
Key words: Radar interferometry; Europe; Numerical modelling; Earthquake ground mo-
tions; Earthquake hazards; Earthquake source observations; Seismic attenuation.
1 INTRODUCTION
An earthquake of Mw6.8 occurred in the Elazı ˘
gregionofeastern
Turkey on 24 January 2020 at 20:55 local time (17:55 UTC), causing
loss of life and severe damage in the epicentral area. According to
the information provided by the Earthquake Department of the Dis-
aster and Emergency Management Presidency, AFAD, there were
46 reported fatalities and over 1600 injuries in Elazı˘
g, Malatya
and Diyarbakir. There are an estimated 10000 people homeless at
this time, sheltering in containers, tents and public refuge sites in
schools, sports facilities and dorms. The earthquake was reported
to be on a segment of the ∼580-km-long left-lateral continental
strike-slip East Anatolian Fault Zone (EAFZ), which is one of the
two major active strike-slip fault systems in Turkey, other being
the ∼1500-km-long right-lateral strike-slip North Anatolian Fault
Zone (NAFZ, inset Fig. 1). The earthquake epicentre is provided by
the different national and international Institutes (including AFAD,
KOERI, USGS, INGV, GCMT, CPPT, ERD, IPGP, GFZ and EMSC)
with rather different locations (in this work we use both the hypocen-
tre information of the main shock based on solution from AFAD and
from the Kandilli Observatory and Earthquake Research Institute,
KOERI, respectively); however, the magnitude has been assessed
1054 C
The Author(s) 2020. Published by Oxford University Press on behalf of The Royal Astronomical Society.
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
The 24 January 2020, Mw 6.8 Elazi˘
g earthquake 1055
Figure 1. Seismotectonic framework of the study area. The solid lines represent the main fault segments of the East Anatolian Fault Zone (EAFZ) with labelled
name: (1) Amanos, (2) Pazarcik, (3) Erkenek, (4) P¨
ut¨
urge, (5) Palu and (6) Ilica fault segments, respectively (after Duman & Emre 2013); the red line is the
fault segment activated during the 2020 Elazı˘
g seismic sequence; the dashed lines are the related fault jogs. Seismicity: the green circles are the aftershocks
of the first 3 months after the main shock (available at https://deprem.afad.gov.tr); the red and white star represent the location of the main shock provided by
AFAD and KOERI, respectively, and its moment tensor solution (U.S. Geological Survey 2020); the grey and yellow stars are the location of the source of the
major (M>6.6) historical and instrumental earthquakes, respectively, in the region with labelled events date (Ambraseys 1989; Ambraseys & Jackson 1998;
Kondorskaya & Ulomov 1999) and their moment tensor solutions (Taymaz et al. 1991). The white arrows represent the direction of plate motion. The dashed
box highlights the area of Fig. 4. The main map is displayed in an oblique Mercator projection with the equator azimuth parallel to the trend of the EAFZ.
The inset shows the main fault systems in and around Turkey (modified from Cetin et al. 2003): NAFZ is the North Anatolian Fault Zone; EAFZ is the East
Anatolian Fault Zone. Arrows indicate relative plate motions. The dashed box is the area of the main figure.
as Mw6.7 or 6.8. Its epicentre was located in the Elazı˘
g province,
at a distance of about 10–20 km WSW of the Lake Hazar (Fig. 1).
Its focal depth ranged from 8 to 23 km. The focal mechanism so-
lution (Fig. 1) indicated that the earthquake is in agreement with
the activation of a ENE–WSW striking left-lateral strike-slip fault.
The affected area is still experiencing repeated aftershocks, with
over 1200 events of magnitude greater than 2.0 within 4 months; 31
aftershocks have been equal to or larger than M4.0. The largestafter-
shocks occurred on 25 January, 19 March and 5 June 2020, and their
magnitudes have been assessed as Mw5.1, 5.0 and 5.0, respectively
(earthquakes solutions available at https://deprem.afad.gov.tr). Af-
tershocks spread along a 50–60 km stretch of the EAFZ between
Sivrice to P¨
ut¨
urge, both ENE and WSW of the hypocentre, spread-
ing outward from the 30 km section of fault that ruptured on 24
January 2020 (Fig. 1).
In this study, great effort has been directed towards understand-
ing the characteristics of source and ground motion associated with
the Elazı˘
g seismic sequence. In this respect, we first use Sentinel-
1 Synthetic Aperture Radar (SAR) data to investigate the ground
displacement field and to infer, by using elastic dislocation mod-
elling, the fault geometry and slip distribution of the causative fault
segment. Then, we performed simulations aiming at reproducing
the high-frequency portion of the strong ground motion records
obtained during 2020 Elazı˘
g earthquake using the stochastic finite-
fault simulation method based on a dynamic corner frequency ap-
proach (Motazedian & Atkinson 2005; Boore 2009). The earth-
quake’s complex nature and the sparse strong motion network in
the epicentral area makes difficult to define the characteristics of
ground shaking particularly in the near-source region. The estima-
tion of ground motions becomes therefore necessary and essential
through the earthquake scenarios and several physics-based deter-
ministic, stochastic and hybrid methods (Boore 2003; Motazedian
& Atkinson 2005; Graves & Pitarka 2010;Maiet al. 2010; Irikura
& Miyake 2011;Menaet al. 2012; Akinci et al. 2017; Pitarka et al.
2017,2019).
The strong ground motion records of Elazı˘
g earthquake were
available and provided recently by AFAD that manage the national
strong motion network in Turkey. The ground motion parameters
in terms of PGA and PGV values and recorded seismograms at 43
strong motion stations within a distance of 200 km from the epi-
centre are used in this study (Table 1). The observed peak ground
accelerations (PGA) were 293 and 239 cm s–2, with the highest
maxima recorded by the two strong motion stations closest to the
fault (within 10 km of the rupture). These stations are located
near the epicentre (Sivrice and P¨
ut¨
urge stations, located ∼3km
to the north and ∼6 km to the southeast of the epicentre, respec-
tively), as well as locations that were in the heavily damaged area.
Therefore, the simulated ground motions calculated at the 1066 vir-
tual stations distributed on a regular grid with 5 km spacing are
validated with the observed ground motion parameters in terms
of PGA and PGV values, and then compared with ground mo-
tion prediction equations (GMPEs) for epicentral distances up to
200 km.
Finally, because the stress changes in the region due to this
earthquake may interact with other fault segments of the EAFZ
in the area, in this study, we investigate the static Coulomb
stress changes induced by the main shock, by examining both
the contribution to the observed aftershock triggering during
the 2020 Elazı˘
g sequence and to the adjacent fault segments
loading.
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
1056 D. Cheloni and A. Akinci
Tab le 1. Peak ground accelerations and velocities from the processed recordings of the 2020 Elazı˘
g earthquake, Turkey, on the strong motion stations up to
150 km distances and local site conditions is defined in Eurocode 8 (EC 8: Seismic Design of Buildings).
Station
code Station name
Station
Lat.
Station
Lon.
PGA-NS
(gal)
PGA-EW
(gal)
PGA-Z
(gal)
PGV-NS
(ms’1)
PGV-EW
(ms’1)
Distance
Rjb
VS30 (ms−1) Site
EC8
2308 Sivrice 38.45 39.31 235.78 292.80 178.57 27.81 45.34 2.63 450-B
4404 P¨
ut¨
urge 38.20 38.87 193.59 228.44 110.62 24.82 28.71 6.16 1380-A
2301 Merkez 38.67 39.19 118.92 142.61 66.24 12.35 8.75 21.95 407-B
0204 Gerger 38.03 39.03 94.24 110.40 59.83 17.10 9.72 29.09 555-B
0212 Sincik 38.03 38.62 43.52 38.52 31.83 7.13 4.80 28.64 -
2302 Maden 38.39 39.68 25.55 31.36 22.77 3.61 2.29 33.92 907-A
2104 Ergani 38.26 39.76 26.75 25.61 24.11 3.12 3.87 45.17 -
4401 Merkez 38.35 38.34 73.23 87.63 37.35 7.74 6.91 39.96 481-B
0205 Kahta 37.79 38.62 25.49 41.01 26.02 6.81 7.38 52.34 660-B
0207 C¸ elikhan 38.03 38.25 32.94 30.50 18.18 5.06 4.27 63.36 660-B
2304 Kovancılar 38.72 39.86 8.823 13.74 5.81 2.05 2.13 67.57 489.B
4412 Yazıhan 38.60 38.18 21.46 18.97 14.44 4.41 4.43 62.90 -
4407 Arguvan 38.78 38.26 31.04 23.27 20.13 3.00 4.55 70.09 735-B
2307 Palu 38.70 39.93 12.81 20.82 11.88 3.74 3.00 67.57 329-C
2105 Dicle 38.36 40.07 10.37 11.09 8.91 1.23 1.70 68.69 -
6201 Merkez 39.07 39.53 11.99 9.70 9.91 1.26 1.21 70.39 -
0210 Merkez 37.77 38.29 23.04 27.43 18.04 8.27 5.48 70.09 -
4406 Akcada˘
g 38.34 37.97 23.60 24.02 14.06 2.07 4.14 71.16 815-A
0201 Merkez 37.76 38.27 35.9 44.61 35.14 10.08 8.10 71.15 391-B
0209 Samsat 37.58 38.48 70.84 58.44 24.71 4.45 4.80 77.42 -
2305 Beyhan 38.73 40.13 3.64 4.78 4.00 0.99 1.21 80.54 907-A
4408 Do˘
gans¸ehir 38.10 37.89 11.35 16.24 15.55 3.00 2.86 81.16 654-B
2306 Karakocan 38.96 40.04 4.356 5.41 2.89 1.12 1.52 86.63 663-B
4405 Hekimhan 38.81 37.94 11.56 11.59 6.67 4.44 1.68 94.12 579-B
2101 Ba˘
glar 37.93 40.20 24.64 26.38 13.64 2.23 2.75 97.07 519-B
2409 Kemaliye 39.28 38.49 13.71 20.43 7.04 2.79 1.73 109.83 875-A
0213 Tut 37.80 37.93 34.95 30.86 14.43 6.04 6.86 91.63 -
6304 Bozova 37.37 38.51 49.02 77.71 29.94 3.60 3.17 100.24 376-B
2415 ˙
Ilic¸ 39.46 38.55 11.30 12.26 8.99 3.13 1.92 124.97 444-B
4410 Kuluncak 38.87 37.68 13.30 15.15 6.64 2.69 1.51 116.20 -
2408 Kemah 39.60 39.03 12.08 16.46 12.08 2.69 2.17 126.10 416-B
2106 Lice 38.46 40.65 8.71 10.57 8.71 1.08 1.30 118.55 -
4409 Darende 38.56 37.49 7.91 6.95 6.68 1.03 1.24 117.11 -
0208 Golbasi 37.79 37.65 18.06 12.42 7.62 3.38 4.31 113.26 469-B
6302 Virans¸ehır 37.23 39.75 16.93 14.05 9.39 1.74 1.81 136.26 936-A
6202 P¨
ul¨
um¨
ur 39.49 39.90 8.26 6.54 3.66 1.46 0.99 125.04 -
2412 C¸a˘
glayan 39.59 39.69 2.56 2.43 1.70 0.52 0.74 129.21 955-A
1213 Adakli 39.23 40.48 8.26 6.54 3.66 2.08 1.14 149.50 -
2 TECTONIC SETTING,
SEISMOTECTONIC AND SEISMICITY
Two major faults contribute to the majority of the seismic hazard in
the region, the North Anatolian Fault Zone (NAFZ) and the Eastern
Anatolian Fault Zone (EAFZ, Fig. 1). Many large earthquakes have
occurred along these faults (Ambraseys 1989) as a result of the
ongoing movement between the Eurasian, African, Arabian and
Anatolian plates. The NAFZ with right-lateral faulting is extending
from Istanbul in the west to Karlıova in the east. During the twentieth
century this fault zone has produced many large earthquakes with
surface rupturing and with a westward migrating sequence (Barka
& Kandinsky-Cade 1988). Around the Karlıova region, NAFZ joins
the SW-trending EAFZ.
The EAFZ forms a ∼580 km left-lateral strike-slip transform
boundary between the northward moving Arabian Plate and west-
ward moving Anatolian block (Fig. 1), resulting in a left-lateral
slip rate of ∼10 ±1mmyr
–1 on the EAFZ (e.g. McClusky et al.
2000; Reilinger et al. 2006; Aktug et al. 2016), but its faulting
is less continuous and less localized than that of the NAFZ (Am-
braseys 2009). In fact, the EAFZ constitutes a complex left-lateral
strike-slip fault zone and it is divided into a number of fault seg-
ments (Fig. 1), as suggested by different authors based on variations
in trends, location of geometric discontinuities, extent of surface
ruptures, stepovers and bend structures along the EAFZ, mapping
of active faults, seismological and palaeoseismological data (e.g.
Hempton et al. 1981; Barka & Kadinsky-Cade 1988;Sarogluet al.
1992; Herece 2008; Duman & Emre 2013). According to Duman &
Emre (2013), from SW to NE, the fault segments of the main EAFZ
fault strand are: the Amanos, Pazarcık, Erkenek, P¨
ut¨
urge, Palu, Ilıca
and Karlıova segments (Fig. 1). The length of these fault segments
varies from 31 to 112 km, while their strikes vary from N35◦Eto
N75◦E (Duman & Emre 2013).
Since this zone is tectonically very active, a series of large sur-
face rupturing earthquakes occurred along the EAFZ main strand
during the last centuries (Fig. 1). From NE to SW the following
large earthquakes have occurred: the 1866 earthquake of Ms7.0
that can be correlated to the Karlıova segment (Ambraseys & Jack-
son 1998); the 1971 Bing¨
ol earthquake of Ms6.8 occurred between
Karlıova and Bing¨
ol (McKenzie 1972); the 1874 earthquake of Ms
7.1 occurred on the Palu segment (Ambraseys & Jackson 1998;
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
The 24 January 2020, Mw 6.8 Elazi˘
g earthquake 1057
Figure 2. (a) Data, (b) model and (c) residual sampled points from the
unwrapped interferogram showing the coseismic displacement field from
the Sentinel-1 ascending track 116. The pink star indicates the main shock
epicentre provided by AFAD.
Figure 3. (a) Data, (b) model and (c) residual sampled points from the
unwrapped interferogram showing the coseismic displacement field from
the Sentinel-1 descending track 123. The pink star indicates the main shock
epicentre provided by AFAD.
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
1058 D. Cheloni and A. Akinci
Figure 4. (a) Coseismic slip distribution on the P¨
ut¨
urge segment of the EAFZ. The contouring (in cm) indicates the major retrieved coseismic patches of slip.
The pink and white stars are the AFAD and KOERI locations of the main shock, respectively; the white blue stars are the major aftershocks with M>5. The
red beach ball indicates the mechanisms of the main shock, while the green beach balls are the M>4 aftershocks that occurred in the first 3 months from
the main shock. Green circles are aftershocks between 23 January and 23 April. The red box represents our best-fitting uniform slip solution. (b) Estimated
slip distribution as a function of depth (symbols as in the top panel). The white blue and green stars are, respectively, the major aftershocks with M>5and
M>4. Note that many hypocentres were located at a fixed depth of 7 km from the automatic AFAD location.
Cetin et al. 2003); the 1875 (Ms6.7) that might have been gen-
erated along the easternmost termination of the P ¨
ut¨
urge segment
(Ambraseys 1989; Cetin et al. 2003) or within the complex re-
leasing bend geometry that is inferred to exist within Lake Hazar
(Duman & Emre 2013); the 1905 (Ms6.8) earthquake occurred on
the Yarpuzlu restraining bend which is located at the western tip of
the P¨
ut¨
urge segment (Ambraseys 1989); the 1893 earthquake of Ms
7.2 occurred on the Erkenek segment (Ambraseys & Jackson 1998);
the 1513 earthquake of Ms7.4 has been attributed to the Pazarcık
segment (Herece 2008); the 1822 earthquake of M7.5 that might
have been generated on the Amonos fault segment (Ambraseys &
Jackson 1998;Seyreket al. 2007).
Prior to the 24 January 2020 Elazı˘
g earthquake sequence, taking
account of the time elapsed from the last event, the slip rate, seis-
mological and palaeoseismological data, some authors have iden-
tified some important seismic gaps along the main strand of the
EAFZ, the ∼80-km-long Pazarcık (Nalbant et al. 2002; Karaba-
cak et al. 2011), the ∼100-km-long Amanos and the ∼95-km-
long P¨
ut¨
urge segments (Duman & Emre 2013; Aktug et al. 2016,
Fig. 1). According to these authors these fault segments have
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
The 24 January 2020, Mw 6.8 Elazi˘
g earthquake 1059
Figure 5. Slip distribution model on the P¨
ut¨
urge fault segment estimated in this study which is used in stochastic ground motion simulations. The pink and
white stars represent the location of the main shock provided by AFAD and KOERI, respectively; the white blue and green stars are, respectively, the major
aftershocks with M>5andM>4 projected on the fault surface (within 3 km). The grey arrows indicate the estimated slip directions for each subfault.
Tab le 2. Model parameters related to source, path and site terms that are considered for the ground motion simulations for the Mw6.8
Elazı˘
g earthquake.
Parameters Values Reference
Fault (strike and dip) 242◦–75◦This study
Fault dimension 51 ×24 km2This study
Moment magnitude 6.8 This study
Depth of the top of fault plane 0.0 km This study
Subfault dimension 1.5 ×1.5 km2This study
Stress drop 9 This study
Crustal shear wave velocity (β)3.5kms
–1 G¨
ok et al. (2007)
Crustal density 2800 kg m–3
Rupture velocity 0.8×β
Pulsing area percentage 50 per cent Boore (2009)
Kappa parameter 0.035 and 0.04 s Boore & Joyner (1997)
NEHRP generic rock and generic soil site BSSC 2001
Distance dependent of duration 0.0 (0–10 km) Atkinson & Boore (1995)
0.1 (R>10 km)
Attenuation model, Q(f) 100f0.43 Akinci et al. (2014)
Geometrical spreading Coef. r−1.0 r≤100 km Akinci & Antonioli (2013)
r−0.5 r≤100 km
Window function Saragoni Hart Boore (1983)
Local amplification NEHRP sites Boore & Joyner (1997)
therefore the potential to produce destructive earthquakes in the
future.
3 SENTINEL-1 SAR DATA
We use SAR data acquired by the Sentinel-1 satellites in TOPS (Ter-
rain Observation by Progressive Scans) mode, exploiting two as-
cending and two descending interferograms to measure the ground
displacement due to the 24 January, 2020, Mw6.8 Elazı˘
g earthquake.
The epicentral area is in fact covered by four Sentinel-1 tracks: the
ascending tracks 116 and 043 (Figs 2, S1 and S2), and the descend-
ing tracks 123 and 021 (Figs 3, S3 and S4). The ascending track
116 and the descending track 021 had the last pre-earthquake image
acquisitions on 21 January, while the ascending track 043 and the
descending track 123 on 22 January. Subsequent acquisitions were
made at 6-d interval. To reduce the contribution from potential post-
seismic deformation, we form the coseismic interferograms using
images acquired closest to the earthquake (the first post-seismic ac-
quisitions along the ascending track 116 and the descending track
021 were on 27 January; while for the other tracks were on 28
January).
We process the data using the Sentinel Application Platform
SNAP software that is provided freely by the European Space
Agency. We use precise orbit data and Shutter Radar Topography
Mission 1 arcsec data (Jarvis et al. 2008) for the image removal
of flat-earth phase and topographic phase. The resulting interfero-
grams were then filtered by applying the Goldstein filter (Goldstein
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
1060 D. Cheloni and A. Akinci
Figure 6. Residuals between observed and simulated PGA and PGVs calcu-
lated for sixteen drop parameters, ranging from 5 to 20 MPa, that minimize
the misfit between observed and simulated ground motion.
0.0
0.1
0.2
0.3
0.4
0.5
P
G
A
g
0.0
0.1
0.2
0.3
0.4
0.5
P
G
V
m
/
s
38˚00'
38˚00'
38˚30'
38˚30'
39˚00'
39˚00'
39˚30'
39˚30'
40˚00'
40˚00'
38˚00' 38˚00'
38˚30' 38˚30'
39˚00' 39˚00'
0.1
0.1
0.2
0.3
0.4
0.5
38˚00'
38˚00'
38˚30'
38˚30'
39˚00'
39˚00'
39˚30'
39˚30'
40˚00'
40˚00'
38˚00' 38˚00'
38˚30' 38˚30'
39˚00' 39˚00'
0 50
km
2308
4404
0204
0212
2302
4401
0205
0207
4212
2304
0210
0201
230
4
2301
2104
Malatya
Elazig
38˚00'
38˚00'
38˚30'
38˚30'
39˚00'
39˚00'
39˚30'
39˚30'
40˚00'
40˚00'
38˚00' 38˚00'
38˚30' 38˚30'
39˚00' 39˚00'
0.1
0.1
0.1
0.2
0.3
0.4
0.5
38˚00'
38˚00'
38˚30'
38˚30'
39˚00'
39˚00'
39˚30'
39˚30'
40˚00'
40˚00'
38˚00' 38˚00'
38˚30' 38˚30'
39˚00' 39˚00'
0 50
km
2308
4404
0204
0212
2302
4401
0205
0207
4212
2304
0210
0201
230
4
2301
2104
Malatya
Elazig
Figure 7. Spatial distributions of synthetics using the spectral parameters
giveninTable2in terms of (a) PGA and (b) PGV values, calculated at rock
site (BC type soil classification, VS30 =760 m s−1), respectively. The region
is divided into regular grid spacing of 5 km as indicated by small-black dots
shown in figure so that the simulations are performed for the 1066 virtual
stations. The grey box represents our extended fault plane solution. The pink
star indicates the main shock epicentre provided by AFAD.
&Werner1998) to reduce interferometric phase noise. Finally, we
unwrapped the phase using Statistical-Cost Network-Flow Algo-
rithm for Phase Unwrapping (SNAPHU) (Chen & Zebker 2001)and
the interferograms were geocoded to obtain the ground deforma-
tion maps. Because of the intrinsic ambiguity of phase unwrapping,
similarly to Wang & Burgmann (2020), we flatten the unwrapped
interferograms by fitting a polynomial function to phase at pixels
away from the epicentre, where the expected ground deformation is
small. Unfortunately, there are no permanent GNSS (Global Nav-
igation Satellite System) stations in the near-field of the epicentre
for a direct comparison with InSAR Line-Of-Sight (LOS) displace-
ments. In fact, the nearest GNSS station is located about 35 km
from the epicentre, in Elazı˘
g city, and has a significant static off-
set of only ∼3 cm towards SW (available in an open file report
at https://deprem.afad.gov.tr/depremdokumanlari/1831). Neverthe-
less, the InSAR displacements are in good agreement with Elazı˘
g
GNSS measurements projected onto the LOS.
The ground deformation retrieved from the ascending and de-
scending unwrapped interferograms are characterized by two ENE–
WSW striking deformation lobes located on both sides of the EAFZ
alignment, with maximum LOS displacement of about 25–30 cm
(Figs 2and 3). The observed differences in the ascending and
descending LOS maps reveal a combination of mainly horizontal
movements consistently with the strike-slip left-lateral mechanisms
of the EAFZ.
4 FAULT GEOMETRY AND COSEISMIC
SLIP
In order to image the fault geometry and slip distribution of the
2020 Elazı˘
g main shock, we performed fault slip modelling using
rectangular dislocations embedded in an elastic, homogeneous and
isotropic half-space (Okada 1985), following a standard two-steps
procedure (e.g. Cheloni et al. 2010,2019): (1) we inverted the LOS
displacements to retrieve the fault geometry and then (2) the best-
fitting uniform-slip fault parameters are used as apriorifor the
estimation of the coseismic slip distribution. Before modelling, the
InSAR interferograms were down-sampled using a resolution-based
down-sampling scheme (Lohman & Simons 2005,FigsS1,S2,S3
and S4).
In the first step, we carried out a non-linear optimization of the
fault geometry by using a simulated annealing algorithm (Corana
et al. 1987). The best-fitting uniform slip model is described by
a 242◦ENE–WSW striking and 75◦NW dipping strike-slip (rake
about –7.5◦) 32.1 km ×8.5 km fault plane passing through the
hypocentral location (Fig. 4, red box) and in good agreement with
focal solutions. The average uniform slip is 1.3 m, which using a
value of 30 GPA for rigidity, yields an estimated seismic moment
of 1.04 ×1019 Nm, equivalent to a Mw6.7 earthquake (the results
of the uniform-slip model are displayed in Figs S5–S8).
In the second step, in order to estimate slip distribution on the
fault plane, we extended the uniform slip fault to capture the area
affected by aftershocks and subdivided the fault into small patches
of constant size (1.5 km ×1.5 km). We apply positivity constraints
and regularize the linear inversion by applying spatial smoothing
(Fig. S9). Additional terms consisting of a linear ramp for each
InSAR interferograms are also included in the inversion and relative
weights were applied to properly combine the different data sets
(Fig. S10).
The best-fitting slip distribution on the extended fault plane
(51 km ×24 km) agrees with the distribution of aftershocks (Figs 4
and 5). The fit to the data significantly improves passing from the
uniform slip to a variable slip model: from 2.75 to 1.64 cm and from
2.47 to 1.78 cm for the LOS displacements on ascending track 116
and descending track 123, respectively, and from 3.22 to 1.65 cm
and from 3.14 to 1.29 cm for the LOS displacements on ascending
track 043 and descending track 021, respectively (Figs 2,3,S11
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
The 24 January 2020, Mw 6.8 Elazi˘
g earthquake 1061
Figure 8. Horizontal component acceleration, and velocity time history plots at three closest stations to the fault rupture. Recorded (left-hand panels, black
and red colour for two horizontal components EW and NS, respectively) and simulated (right panels) at the closest Sivrice, P¨
ut¨
urge, and Gerger sites to the
earthquake rupture for the 2020 Elazı˘
gMw6.8 earthquake.
and S12). The coseismic slip distribution model shows two major
asperities with peak slip of about 2–2.3 m, located WSW respect
to the epicentre, and mostly confined within the first 8 km, that
is roughly contained in the uniform slip fault (red box in Fig. 4),
for a length of about 40 km, thus leaving unbroken the WSW part
of the P¨
ut¨
urge fault segment. In addition, our variable slip model
shows some slip also to the ENE of the epicentre, in an area where a
number of M>4 earthquakes occurred. The resulting total seismic
moment (1.70 ×1019 Nm) agrees with an Mw6.8 earthquake. The
rake angles of the major coseismic fault patches are consistent with
a predominantly pure left-lateral strike-slip faulting mechanisms.
The two main asperities characterizing our slip model agree with
the up-dip rupture episodes and with the unilaterally WSW rupture
directivity, as retrieved by USGS finite fault analysis (U.S. Geolog-
ical Survey 2020) and by the recent study of Melgar et al. (2020).
Our retrieved slip located ENE of the epicentre, implies instead
bilateral rupture propagation.
The inversion was repeated using separately the ascending and
the descending tracks, respectively. The resulting coseismic slip
distribution is showed in Fig. S13. They are quite similar, showing
two major coseismic slip patches located WSW of the hypocentre,
suggesting these patches to be a robust feature of the retrieved slip
distribution. Notwithstanding, there are some differences, in fact,
the descending tracks-only inversion (Fig. S13b) appears to require
a more eastward position of the smaller patch of slip located ENE
of the epicentre as retrieved in the joint inversion.
5 STOCHASTIC MODELLING OF
HIGH-FREQUENCY GROUND MOTION
In this section, we attempt to simulate the high-frequency ground
motions for the Mw6.8 Elazı˘
g earthquake using a well-known
stochastic finite-fault simulation method (Motazedian & Atkinson
2005). To do so we gather several parameters related to the source,
the path and the site that are essential and requested by the adopted
approach have been chosen among those several existing models
published for the eastern Turkey.
5.1 Finite-fault source model
The necessary source parameters are defined in terms of the fault
geometry, rupture velocity, stress drop and seismic moment. The slip
distribution along the fault plane is another crucial input parameter
particularly for the ground motion simulation in the source area.
The slip model determined in this study is considered as an input
for our ground motion calculations. The fault plane geometry was
determined from the geodetic inversion with strike 242◦and dip
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
1062 D. Cheloni and A. Akinci
Figure 9. Comparison of observed and simulated Fourier amplitude spectra of acceleration (cm s−1) at six selected stations. Simulated spectra (solid blue
lines) and observed spectra for the two horizontal EW and NS components (solid red and black lines, respectively).
75◦and with the top of the fault plane at 0 km depth (Fig. 5). Two
patches of slip are examined over the fault plane: the larger one is
located at the centre and has the highest slip of about 2.3 m and the
other is located at southwestern part of the fault and has about 2 m
of slip. The dimensions of the modelled left-lateral strike-slip fault
plane are 50 km ×25 km and our rupture model includes 34 ×16
subfaults.
There have been several efforts to determine the fault rupture and
mechanism of the 2020 Elazı˘
g earthquake, established on various
data set (teleseismic, GPS and InSAR) and alternative inversion
methods (e.g. Melgar et al. 2020). Our inversion results agree with
these models defining the location of the slip asperities on the
fault plane and the quantitative description of the slip during the
earthquake rupture. The parameters of the finite-fault source model
used in our ground motion simulations are listed in Table 2.
5.2 High-frequency seismic wave attenuation model
The seismic wave propagation and the seismic attenuation is an
important topic and it is essential for the prediction of earthquake
ground motion in seismic hazard analysis. Determining the mech-
anism of attenuation helps to understand the regional differences
observed on the ground motion. There have been several studies
to determine the attenuation characteristics in eastern Turkey using
various database and methods (Mitchell et al. 1997; Mitchell &
Cong 1998;Zoret al. 2007; Pasyanos et al. 2009; Sertcelik 2012;
Akinci et al. 2014).
Recently Akinci et al. (2014) provided a complete description of
the characteristics of the source and the attenuation of the ground
motion around the Lake Van region (eastern Turkey) using a large
set of broadband data of the main shock and aftershocks of the 23
October 2011 Mw7.1 Van earthquake. They observed strong crustal
attenuation, Q(f)=100f0.43 together with the geometrical spreading,
g(r), occurring at a hypocentral distance of 40 km which it changes
from a body-wave-like function g (r)∝r−1.0 to a functional form
g(r)∝r−0.5 expected for surface waves. These results are not very
different from those observed by Zor et al. (2007)usingtheLgwaves
through the two stations method and back projection tomography;
QLg0 was around 100 with its frequency dependence n=0.4–0.6.
Sertcelik (2012) has also estimated the seismic wave attenuation
using the coda waves as a function of the lapse time and frequency
over the EAFZ. Although those studies result with a similar Qvalue
(∼100) over the Eastern Anatolia, in our study, we decide to use
most recently characterized seismic attenuation parameters as given
by Akinci et al. (2014)(Table2).
The geometrical spreading values governed by the crustal struc-
ture, capable of producing post-critical reflections from mid crustal
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
The 24 January 2020, Mw 6.8 Elazi˘
g earthquake 1063
Figure 10. Comparisons between observed and simulated (a) peak ground
acceleration, PGA and (b) peak ground velocity, PGV. The simulated values
are obtained using a dynamic corner frequency approach. The PGA and
PGV simulated values were estimated for an earthquake of Mw6.8 using
the parameters given in Table 2and the proposed fault geometry and slip
distribution obtained in this study, for the 1066 virtual stations (light blue
triangles). The black crosses represent the observed data of Elazı˘
g earth-
quake recorded in two horizontal components (EW light grey and NS black
coloured crosses) at 43 stations while the red triangles represent those from
simulated seismograms at the same stations on a rock site, Vs30 =760 ms−1.
Ground Motion Predictions Equations (GMPEs), obtained using Boore et al.
(2014) (short dashed line); Bindi et al. (2014) (long dashed line) and Akkar
& Cagnan (2010) (thick solid line) and the ±1σthe total standard deviation
of AC10 (thin solid lines) for a rock site of Vs30 =760 m s−1are also shown.
and Moho discontinuities are also appropriated as given by Akinci
& Antonioli (2013). The stress drop parameter σ which governs
the levels of the acceleration spectrum at high frequencies is esti-
mated between 8 and 20 MPa by Akinci et al. (2006), Malagnini
et al. (2010), Akinci & Antonioli (2013) and Akinci et al. (2014)
for the 1999 Kocaeli, Mw7.2 earthquake with strike-slip faulting
and for the 2011 Van Lake Mw7.1 earthquake with reverse faulting.
Since the stress drop parameter is not estimated specifically for the
Mw6.8 Elazı˘
g earthquake, we calculate the residuals between the
observed PGA and PGV values and those from simulations over
16 different stress drop parameters, from 5 to 20 MPa. Finally, we
select the σ that minimizes the misfit between the observed and
simulated PGA and PGV data.
Results presented in Fig. 6demonstrate how the choice of the
stress drop parameter affects the misfit. The lowest bias determined
by averaging the residuals over the 43 stations indicates the best
parameter that could be considered for the simulations. The chosen
stress drop parameter for our ground motion estimations that pro-
duces a good fit to the observed data for the PGA and PGV is equal
to σ =9MPa.
6 HIGH FREQUANCY GROUND
MOTION SIMULATIONS FOR THE Mw
6.8 ELAZIG EARTHQUAKE
6.1 Spatial distribution of simulated ground motions
In order to investigate the spatial variation of ground motion caused
by the Mw6.8 Elazı˘
g earthquake, we calculate ground motion pa-
rameters together with the synthetic time histories at 43 recording
stations, and at 1066 virtual stations on a regular grid with 5 km
spacing, covering a regionbetween 38.25–39.5◦E and 38.0–38.75◦N
up to 200 km distances. The high frequency seismograms are gen-
erated using a stochastic finite-fault model approach, based on a
dynamic corner frequency (Motazedian & Atkinson 2005; Boore
2009) and considering the spectral parameters as given in Table 2.
Spatial distribution of the estimated ground motion parameters in
terms of PGA and PGV values within the study area is shown in
Figs 7(a) and (b), respectively.
The site-amplification effect is considered as uniform for the
whole region and referred to the BC type generic rock site condition
(Vs30 =760 m/s, the shear wave velocity averaged over the top 30 m
of the soil, BSSC 2001). So that, the spatial distributions of PGA
and PGV values mainly reflects the source effects. We observe that
the largest ground shaking is concentrated along the rupture fault
plane, where PGA and PGV values raising up to 0.5 g and 40 cm/s,
respectively. Particularly, we observe the strongest ground shaking
within the surface projection of the fault around the location of the
two slip asperities; one asperity is located at the centre of the fault
plane and is larger and much stronger than the second asperity that
is located in the southern section of the fault plane. Finally, while
the near-field results are governed by the source effects, such as the
distribution of asperities on the fault plane, intermediate distances
are controlled mainly by the propagation and the attenuation of
seismic waves with distance. However, since the stochastic approach
adopted in our study is not very sensitive to large-scale source
related directivity effects, the simulated ground motion may not be
suitable for producing the expected ground motion variability in the
near-fault region.
Some of the simulated synthetic waveforms together with the reg-
istered seismograms are shown in Fig. 8at three near-fault stations
(Sivirce, P¨
ut¨
urge and Gerger). The soil conditions are provided for
Sivrice, P¨
ut¨
urge and Gerger stations in the AFAD website. Station
Sivrice (2304) located on the soil type with Vs30 ∼450 m s−1is the
closest station to the fault rupture, with Joyner and Boore distance
to the fault surface of 3 km. The simulated PGA and PGV values are
328 cm s–2 and31cms
–1, computed considering the soft soil site am-
plifications, while the observed PGA values are 238 and 293 cm s–2
and the observed PGV values are 27 and 45 cm s–1 for the two
horizontal components, respectively. Simulations at P ¨
ut¨
urge station
(2305) located on the rock site with a Vs30 ∼1380 m s−1and at 6 km
distance from the fault surface, result in a simulated PGA value of
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
1064 D. Cheloni and A. Akinci
Figure 11. The spectral accelerations from the recorded (thin lines) and simulated (thick lines) seismograms for Sivrice, P¨
ut¨
urge and one for a virtual station
(red colour) located close to the nucleation point where maximum slip is released over the fault rupture, and comparison with the code-based spectrum from
old DBYBHY2007 and new TBDY2018 building codes, respectively.
268 cm s–2 while the observed values are 228 and 193 cm s–2 .The
simulated PGV value is instead 25 cm s–1, while the observed values
are 29 and 25 cm s–1 for the two horizontal components (NS and
EW, respectively). The site condition at the Gerger (0204) station is
given as Vs30 =550 ms−1; our simulated ground motion parameters
at this site are 87.86 cm s–2 for PGA and 10.60 cm s–1 for PGV val-
ues, respectively, while the recorded values are 110.1–94.24 cm s–2
and 17.1–9.71 cm s–1 for two horizontal components, respectively.
Therefore, the PGA and PGV parameters are very well reproduced
by the stochastic modelling approach for short distances.
In order to show the effectiveness of our simulations to reproduce
observations in frequency domain, in Fig. 9we compare the Fourier
amplitude spectra with the recorded ones at six selected strong
ground motion stations. As shown in Fig. 9, simulations provide
modest estimates of the general shape and amplitudes of the spectra
for almost all of the stations. The misfit at P¨
ut¨
urge and Gerger sta-
tions, at the higher frequencies could be attributed to inaccurate site
amplification function and κparameter adopted for those stations in
our study. The synthetic spectra calculated using a high-frequency
attenuation parameter κ=0.035 s and the frequency-dependent site
amplification of the BC type site classification may be insufficient
to capture all the features of the spectral variation and real transfer
function driven by the detailed velocity-depth profile (Vs30).
6.2 Comparison of observed and simulated ground
motions with selected GMPEs
We validate our simulated ground motion parameters against the
43 observed PGA and PGV values up to 200 km distances from
the recordings of the Elazı˘
g earthquake (Table 1) and compare
our simulations with the three selected GMPEs derived for the
active shallow crustal regions. These include the GMPEs developed
(1) within the context of the Next Generation Attenuation (NGA)
models given by Boore et al. (2014) (hereafter, BSSA14); (2) from
European and the Middle East strong motion database of Bindi
et al. (2014) (hereafter, BIN14) and (3) from national strong motion
database and events as the Turkish attenuation model of Akkar &
Cagnan (2010, hereafter AC10).
In Fig. 10 the two horizontal components of the PGAs recorded
by the total of 43 strong motion stations of Turkish network are
plotted up to 200 km as a function of the distance, and compared
with the values calculated from the two GMPEs, BSSA14, BIN14
and the AC10, together with our simulated PGAs. All GMPEs are
derived for strike-slip faulting style and rock conditions, Vs30 =760
ms
−1. Simulations are also performed for the BC type site class,
Vs30 =760 m s−1and κ=0.035 s for all the sites (Boore & Joyner
1997; BSSC 2001). As can be seen from Fig. 10(a) the observed
PGAs are mostly overestimated both by BSSA14 and BIN14 at all
distances, while the AC10 model offers better fit to the observed
and the simulated data at the rock sites. At distances greater than
100 km, the recorded PGAs decay with distance is larger than that
predicted by the GMPEs.
The three GMPE models are in good agreement with the simu-
lated and recorded PGV data at short distances although the BIN14
slightly overestimates both the recorded and the simulated data.
Moreover, the observed PGVs are much scattered than the PGAs
and makes ambitious to confirm the better fit with the predictions
particularly at larger distances. It is interesting to note that the
simulated PGA and PGV values determined using region specific
parameters and regional strong motion data are in good agreement
both with those estimated from the empirical Turkish GMPE model
of AC10 and the simulated data at all distances.
In Fig. 11, we compare the spectral accelerations from the sim-
ulated and registered seismograms for Sivrice, P¨
ut¨
urge and from a
virtual station close to the nucleation point (where maximum slip is
released over the fault rupture), with the code-based design spectra
(for Z3, Vs30 >200 m s–1 ,andZC,V
s30 360–760 m s–1 , site classifi-
cation) from both old DBYBHY2007 and new TBDY2018 (T¨
urkiye
Bina Deprem Y¨
onetmeli˘
gi 2018) building codes, respectively. As
it is seen in Fig. 11, response spectra of P¨
ut¨
urge and Sivrice sta-
tion records are well-below the 475 yr design spectra from the new
Turkish building codes (TBDY-2018), differently from the old code
(DBYBHY-2007). The spectral acceleration of the virtual station
(0533) over the fault rupture (where there are no strong motion reg-
istration/recordings available) is also covered by the spectra design
by the new TBDY-2018. It can be observed from Fig. 11 that the
Sivrice’s response spectra has long periods and higher amplitudes
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
The 24 January 2020, Mw 6.8 Elazi˘
g earthquake 1065
Figure 12. Slip distribution and Coulomb failure stress variation produced by the 24 January main shock for left-lateral ENE–WSW striking faults, in
agreement with fault orientations of the EAFZ in this area. (a) Slip distribution: the green circles represent the aftershocks of the Elazig seismic sequence;
pink star is the main shock epicentre provided by AFAD, while grey stars are previous large (M>6.6) historical and instrumental events along the EAFZ.
Solid lines represent the main fault segments, while dashed lines the fault jogs. Abbreviations: Y-RDB, Yarpuzlu restraining double bend; H-RB, lake Hazard
releasing bend. The black box represents our extended fault plane solution. The main map is displayed in an oblique Mercator projection with the equator
azimuth parallel to the trend of the EAFZ as in Fig. 1, (b) Change of Coulomb stress (friction coefficient =0.4) computed at a reference depth of 10 km. The
contouring (black solid lines, in cm) indicated the major coseismic slip distribution of the Elazı˘
g earthquake. The ellipse corresponds to the slip deficit area as
suggested in this study. Other symbols are as in panel (a).
due to its site characteristic when compared to that from the rock
motions recorded at P¨
ut¨
urge.
7 VARIATIONS OF STATIC COULOMB
STRESS IN THE STUDY AREA
It is well established that the coseismic slip causes some varia-
tions in static stress that may trigger subsequent earthquakes as well
as patches of aseismic slip on unbroken portions of the causative
fault itself or/and on adjacent faults (e.g. Lin & Stein 2004). In
this context it is important for hazard assessment in the study area
to evaluate the potential state of stress of the unruptured portions
of the P¨
ut¨
urge segment as well as of adjacent segments follow-
ing the 24 January 2020, Mw6.8 Elazı˘
g main shock. To inves-
tigate such stress variations, we calculated the Coulomb stress
change induced by the main shock using our preferred slip dis-
tribution (Fig. 12a) on fault segments having the same mechanism
and geometry as the main event, and assuming an effective fric-
tion coefficient of 0.4, as is commonly used in stress interaction
studies (e.g. Freed 2005; Akinci & Antonioli 2013). As expected,
we find increased Coulomb stress when projected on ENE–WSW
striking left-lateral slip faults at both terminations of the mod-
elled fault plane; (1) ENE near Lake Hazar and (2) WSW of the
hypocentre, respectively, in areas where also a large number of
aftershocks occurred following the 24 January Elazı˘
g earthquake
(Fig. 12b).
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
1066 D. Cheloni and A. Akinci
As regard the first stressed area, this zone was likely ruptured
during the 1875 earthquake. In fact, the 1875 event has been gen-
erated along the easternmost termination of the P ¨
ut¨
urge segment
(Ambraseys 1989) or within the complex releasing bend geometry
that is inferred to exist within Lake Hazar (Duman & Emre 2013,
H-RB in Fig. 12). Differently, the second stressed area, WSW of the
hypocentre, might be a portion of the P¨
ut¨
urge fault segment that has
not yet been broken by any seismic event in the last centuries (red
ellipse in Fig. 12b). In fact, there is a debate on the exact location
of the 1905 (Ms6.8) earthquake, that is the last relevant seismic
event possibly located around the western end of the P¨
ut¨
urge seg-
ment (Ambraseys 1989). This event might have happened either
on the Yarpuzlu restraining bend (Y-RDB in Fig. 12) or along the
western part of the P¨
ut¨
urge fault segment itself. Depending of the
real location of the 1905 earthquake, the first hypothesis allows the
possibility that the western part of the P¨
ut¨
urge segment may be an
unbroken area that may rupture in a future earthquake(s).
8 CONCLUSIONS
The 24 January 2020 Mw6.8 Elazı˘
g earthquake in Eastern Turkey
was caused by the rupture of a segment of the EAFZ, the (∼75◦)
NNW dipping left-lateral strike-slip P¨
ut¨
urge segment, which has
not ruptured in the recent past and that was considered as a seismic
gap prior of the 2020 seismic sequence (Duman & Emre 2013).
The slip distribution obtained from the geodetic inversion shows
two major asperities with peak slip of ∼2.3 m, located WSW from
the epicentre, thus implying a marked westward directivity for the
Elazı˘
g main shock. The seismic moment release calculated with
the geodetic data is 1.70 ×1019 Nm (equivalent to a Mw6.8),
in agreement with magnitude estimates provided by different na-
tional and international Institutes. The slip distribution along the
∼90-km-long causative P¨
ut¨
urge fault segment implies also that the
western part of the seismogenic fault remained unruptured and was
positively stressed by an increase of Coulomb stress.
We simulate the high-frequency ground motion and seismograms
at 1066 virtual stations ranging between 0.1 and 100 km distances in
the epicentral area to have detailed point of view on ground motion
intensity distribution particularly close to the fault rupture where
the maximum damaged observed. In the near-fault area we observe
that our simulations have a good capability to detect near source
effects and to reproduce the source complexity. The general good
consistency found between synthetic and observed ground motion
both in time and frequency domain, suggests the importance of
the retrieving specific regional seismic parameters. We remark that
ground motion parameters decay faster than the empirical ground
motion equations except that of Turkish GMPEs of AC10 both at
moderate and particularly at larger distance (around 100 km) this
feature is captured by our simulated data. Although the adopted
stochastic approach does not fully produce source directivity ef-
fects, such as coherent pulses in near-fault ground motion, it can be
easily and quickly implemented for both region-specific and path-
specific applications. We also demonstrate that the design spectra
corresponding to 475 yr return period, provided by the new Turkish
building code is not exceeded by the simulated seismograms in the
epicentral area where there are no strong motion stations and no
recordings available.
Finally, our prefer red fault model, computed stress redistribution,
and location of historical earthquakes along the EAFZ highlights
some interesting features that are relevant to seismic hazard assess-
ment in the region. In fact, to the WSW of the 2020 Elazı˘
gseismic
sequence, our fault modelling and stress calculation suggests the
presence of a stressed and possibly unbroken area of the P¨
ut¨
urge
segment that should be considered for future hazard assessment.
For this reason, our results suggest that the occurrence of future
significant earthquakes to the WSW of P¨
ut¨
urge city cannot be ruled
out, and therefore a significant seismic hazard still remains in the
area.
ACKNOWLEDGEMENTS
We would like to thank the Editor, Dr Kosuke Heki, the reviewer Ar-
ben Pitarka of Lawrence Livermore National Laboratory, CA, and
the reviewer Farnaz Kamranzad of University of Tehran, for their
constructive suggestions, which helped to improve the manuscript.
Most of the figures have been created using the Generic Mapping
Tools version 4.2.1 (www.soest.hawwai.edu/gmt) and the software
of Seismic Analysis Code (SAC) is used for many of the calculations
throughout several set of macros. We use Copernicus Sentienel-1
InSAR imagery (https://scihub.copernicus.eu/). Sentinel-1 data are
copyright of Copernicus (2020). We thank everyone at the Earth-
quake Department of the Disaster and Emergency Management
Presidency, AFAD for making the strong motion data available
(https://tadas.afad.gov.tr/).
REFERENCES
Akinci, A, Malagnini, L., Herrmann, R.B. & Kalafat, D., 2014. High-
frequency attenuation in the Lake Van Region, Eastern Turkey, Bull.
seism. Soc. Am., 104(3), doi:10.1785/0120130102.
Akinci, A. & Antonioli, A., 2013. Observations and stochastic modelling
of strong ground motions for the 2011 October 23 Mw 7.1 Van, Turkey,
earthquake, Geophys. J. Int., 192, doi:10.1093/gji.ggs075.
Akinci, A., Aochi, H., Herrero, A., Pischiutta, M. & Karanikas, D., 2017.
Physics-based broadband ground-motion simulations for probable Mw
≥7.0 earthquakes in the Marmara Sea Region (Turkey), Bull. Seismol.
Soc. Am, 107(3), doi:10.1785/0120160096.
Akinci, A., Malagnini, L., Herrmann, R.B., Gok, R. & Sorensen, M.B.,
2006. Ground motion scaling in Marmara region, Turkey, Geophys. J.
Int., 166, 635–651.
Akkar, S. & Cagnan, Z., 2010. A local ground-motion predictive model for
Turkey and its comparison with other regional and global ground-motion
models, Bull. seism. Soc. Am., 100, 2978–2995.
Aktug, B., Ozener, H., Dogru, A., Sanbucu, A., Turgut, B., Halicioglu, K.,
Yilmaz, O. & Havazli, E., 2016. Slip rates and seismic potential on the
East Anatolian Fault System using an improved GPS velocity field, J.
Geod., 94–95, 1–12.
Ambraseys, N., 2009. Earthquakes in the Mediterraneanand Middle East: A
Multidisciplinary Study of Seismicity up to 1900. Cambridge Univ. Press,
968pp.
Ambraseys, N.N., 1989. Temporary seismic quiescence: SE Turkey,
Geophys. J. Int., 96, 311–331.
Ambraseys, N.N. & Jackson, J.A., 1998. Faulting associated with historical
and recent earthquakes in the Eastern Mediterranean region, Geophys. J.
Int., 133, 390–406.
Atkinson, G.M. & Boore, D.M., 1995. Ground-motion relations for eastern
North America, Bull. seism. Soc. Am., 85, 17–30.
Barka, A.A. & Kandinsky-Cade, K., 1988. Strike-slip fault geometry in
Turkey and its influence in earthquake activity, Tectonics, 7, 663–684.
Bindi, D., Massa, M., Luzi, L., Ameri, G., Pacor, F., Puglia, R. & Augliera, P.,
2014. Pan-European ground motion prediction equations for the average
horizontal component of PGA, PGV and 5%-damped PSA at spectral
periods of up to 3.0 s using the RESORCE dataset, Bull. Earthq. Eng.,
12(1), 391–340
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
The 24 January 2020, Mw 6.8 Elazi˘
g earthquake 1067
Boore, D.M., 1983. Stochastic simulation of high-frequency ground motions
based on seismological models of the radiated spectra, Bull. seism. Soc.
Am., 73, 1865–1894.
Boore, D.M., 2003. Simulation of ground motion using the stochastic
method, Pure appl. Geophys., 160, 635–676.
Boore, D.M., 2009. Comparing stochastic point-source and finite-source
ground-motion simulations: SMSIM and EXSIM, Bull. seism. Soc. Am.,
99, doi:10.1785/0120090056.
Boore, D.M. & Joyner, W.B., 1997. Site amplifications for generic rock sites,
Bull. seism. Soc. Am., 87, 327–341.
Boore, D.M., Stewart, P.J., Seyhan, E. & Atkinson, G.M., 2014. NGAWest2
equations for predicting PGA, PGV, and 5% damped PSA for shallow
crustal earthquakes, Earthq. Spectra, 30(3), 1057–1085.
BSSC (Building Seismic Safety Council), 2001. NEHRP recommended pro-
visions for seismic regulations for new buildings, and other structures,
2000 Edition. Part 1: Provisions, Building Seismic Safety Council for the
Federal Emergency Management Agency (Report FEMA368), Washing-
ton, DC, USA.
Cetin, H., Guneyli, H. & Mayer, K., 2003. Paleoseismology of the Palu-Lake
Hazar segment of the East Anatolian Fault Zone, Turkey, Tectonophysics,
374, 163–197.
Cheloni, D. et al., 2010. Coseismic and initial post-seismic slip of the 2009
Mw 6.3 L’Aquila earthquake, Italy, from GPS measurements, Geophys. J.
Int., 181(3), 1539–1546.
Cheloni, D. et al., 2019. Heterogeneous Behavior of the Campotosto Normal
Fault (Central Italy) Imaged by InSAR GPS and Strong-Motion Data:
Insights from the 18 January 2017 Events, Remote Sens., 11, 1482.
Chen, C.W. & Zebker, H.A. 2001. Two-dimensional phase unwrapping with
use of statistical models for cost functions in nonlinear optimization, J.
Opt. Soc. Am., 18(2), 338–351.
Corana, A., Marchesi, M., Martini, C. & Ridella, S., 1987. Minimizing mul-
timodal functions of continuous variables with the “Simulated Annealing”
algorithm, ACM Trans. Math. Softw., 13, 262–280.
Duman, T.Y. & Emre, ¨
O., 2013. The East Anatolian Fault: geometry, seg-
mentation and jog characteristics, Geol. Soc., Lond., Spec. Publ., 372,
495–529.
Freed, T.G., 2005. Earthquake triggering by static, dynamic, and
post-seismic stress transfer, Annu. Rev. Earth Planet Sci., 33,
doi:10.1146/annurev.earth.33.092203.122505.
Goldstein, R.M. & Werner, C.L., 1998. Radar interferograms filtering for
geophysical applications, Geophys. Res. Lett., 25, 4035–4038.
Graves, R. & Pitarka, A., 2010. Refinements to the Graves and Pitarka (2010)
broadband ground motion simulation method, Seismol. Res. Lett., 86(1),
75–80.
G¨
ok, R., Pasyanos, M.E. & Zor, E., 2007. Lithospheric structure of the
continent collision zone: Eastern Turkey, Geophys. J. Int., 169, 3, 1079–
1088.
Hempton, M.R., Dewey, J.F. & Saroglu, F., 1981. The East Anatolian trans-
form fault: along strike variations in geometry and behavior, EOS, Tran.
Am. Geophys. Un., EOS, 62, 393.
Herece, E., 2008. Dogu Anadolu Fayi (DAF) Atlasi. General Directorate of
Mineral Research and Exploration. Special Publications, Ankara, Serial
Number, 14, 359.
Irikura, K. & Miyake, H., 2011. Recipe for predicting strong ground motion
from crustal earthquake scenarios, Pure appl. Geophys., 168, 85–104.
Jarvis, A., Reuter, H.I., Nelson, A. & Guevara, E., 2008. Hole-filled SRTM
for the globe, version4, CGIAR-CSI SRTM 90m Database, available at ht
tp://srtm.csi.cgiar.org.
Karabacak, V., Onder, Y., Altunel, E., Yalciner, C.C., Akyuz, H.S. & Kiyak,
N.G., 2011. Dogu Anadolu Fay Zonunun guney bati uzaniminin paleo-
sismolojisi ve ilk kayma hizi. Proceeding of the Aktif Tektonik Arastirma
Grubu Onbesinci Calistayi (ATAG-15), 19–22 Ekim 2011, Cukurova Uni-
versitesi, Karatas-Adana, 17.
Kondorskaya, N.V. & Ulomov, V.I., 1999. Special catalogue of earthquakes
of the Northern Eurasia (SECNE). http://www.seismo.ethz.ch/static/gsha
p/neurasia/nordasiacat.txt.
Lin, J. & Stein, R.S., 2004. Stress triggering in thrusts and subduction
earthquakes, and stress interaction between the southern San Andreas
and nearby thrust and strike-slip faults, J. geophys. Res., 109(B02303),
doi:10.1029/2003JB002607.
Lohman, R.B. & Simons, M., 2005. Some thoughts on the use of In-
SAR data to constrain models of surface deformation: noise struc-
ture and data downsampling, Geochem. Geophys. Geosyst., 6(Q01007),
doi:10.1029/2004GC000841.
Mai, P.M., Imperatori, W. & Olsen, K.B., 2010. Hybrid broadband ground
motion simulations: combining long-period deterministic synthetics with
high-frequency multiple S-to-S backscattering, Bull. seism. Soc. Am.,
100(5A), 2124–2142.
Malagnini, L., Nielsen, S., Mayeda, K. & Boschi, E., 2010. Energy
radiation from intermediate to large magnitude earthquakes: impli-
cations for dynamic fault weakening, J. geophys. Res., 115(B6),
doi:10.1029/2009JB006786.
McClusky, S. et al., 2000. GPS constraints on plate motions and deformation
in the Eastern Mediterranean: implications for plate dynamics, J. geophys.
Res., 105, 5695–5719.
McKenzie, D.P., 1972. Active tectonics of the Mediterranean region,
Geophys.J.R.astr.Soc.,30, 109–185.
Melgar, D., et al.,, 2020. Rupture kinematics of January 24, 2020
Mw 6.7 Do˘
ganyol-Sivrice, Turkey, earthquake on the East Anatolian
Fault Zone imaged by space geodesy, Geophys. J. Int. , ggaa345,
doi:10.1093/gji/ggaa345 .
Mena, B., Dalguer, L.A. & Mai, P.M., 2012. Pseudodynamic source
characterization for strike slip faulting including stress hetero-
geneity and super-shear ruptures, Bull. seism. Soc. Am., 102(4),
1654–1680.
Mitchell, B. & Cong, L., 1998. Lg coda Q and its relation to the structure
and evolution of continents: a global perspective, Pure appl. Geophys.,
153, 655–663.
Mitchell, B.J., Pan, Y., Xie, J. & Cong, L., 1997. Lg coda Q variation across
Eurasia and its relation to crustal evolution, J. geophys. Res., 102, 22767–
22779.
Motazedian, D. & Atkinson, G.M., 2005. Stochastic finite-fault 800 mod-
elling based on a dynamic 801 corner frequency, Bull. seism. Soc. Am.,
95, 995–1010.
Nalbat, S.S., McCluskey, J., Steacy, S. & Barka, A.A., 2002. Stress accu-
mulation and increased seismic risk in eastern Turkey, Earth. Planet. Sci.
Lett., 195, 291–298.
Okada, Y., 1985. Surface deformation due to shear and tensile faults in a
half-space, Bull. seism. Soc. Am., 75, 1135–1154.
Pasyanos, M.E., Matzel, E.M., Walter, W.R. & Rodgers, A.J., 2009. Broad-
band Lg attenuation modeling in the Middle East, Geophys. J. Int., 177,
1166–1176.
Pitarka, A., Graves, R., Irikura, K., Miyake, H. & Rodgers, A., 2017. Per-
formance of Irikura recipe rupture model generator in earthquake ground
motion simulations with graves and Pitarka hybrid approach, Pure appl.
Geophys., 174(9), doi:10.1007/s00024-017-1504-3.
Pitarka, A., Graves, R., Irikura, K., Miyakoshi, K. & Rodgers, A., 2019.
Kinematic rupture modeling of ground motion from the M7 Kumamoto,
Japan, Earthquake, Pure appl. Geophys., 177, 2199–2221.
Reilinger, R. et al., 2006. GPS constraints on continental deformation in
the Africa-Arabia-Eurasia continental collision zone and implications
for the dynamics of plate interactions, J. geophys. Res., 111(B05411),
doi:10.1029/2005JB004051.
Saroglu, F., Emre, O. & Kuscu, O., 1992. The East Anatolian fault zone of
Tur key , Ann. Tecton., 6, 99–125.
Sertcelik, F., 2012. Estimation of coda wave attenuation in the east Anatolia
fault zone, Turkey, Pure appl. Geophys., 169(7), 1189–1204.
Seyrek, A., Demir, T., Pringle, M.S., Yurtem, S., Westway, R.W.C., Beck,
A. & Rowbotham, G., 2007. Kinematics of the Amos Fault, southern
Turkey, from Ar/Ar dating of offset Pleistocene basalt plates, in Tectonics
of Strike-Slip Restraining and releasing Bends, Vol . 290, pp. 255–284,
eds Cunningham, W.D. & Mann, P., Geological Society London Special
Publications.
Taymaz, T., Eyidogan, H. & Jackson, J., 1991. Source parameters of large
earthquakes in the East Anatolian Fault Zone (Turkey), Geophys. J. Int.,
106, 537–550.
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
1068 D. Cheloni and A. Akinci
T¨
urkiye Bina Deprem Y¨
onetmeli˘
gi (TBDY), 2018. T.C. Ba s¸bakanlık Afet
ve Acil Durum Y¨
onetimi Bas¸kanlı˘
gı, Deprem Dairesi Bas¸kanlı ˘
gı, http:
//www.deprem.afad.gov.tr, Ankara.
U.S. Geological Survey, 2020. Earthquake Hazards Program, available
at: https://earthquake.usgs.gov/earthquakes/eventpage/us60007ewc/; last
accessed on 20 April 2020.
Wang, K. & Burgmann, R., 2020. Co- and postseismic deformation due
to the 2019 Ridgecrest earthquake sequence constrained by Sentinel-
1 and COSMO-SkyMed Data, Seismol. Soc. Lett., 91(4), 1998-2009,
doi:10.1785/0220190299.
Zor, E., Sandvol, E., Xie, J., Turkelli, N., Mitchell, B., Gasanov, A.H. &
Yetirmishli, G., 2007. Crustal attenuation within the Turkish Plateau and
surrounding regions, Bull. seism. Soc. Am., 97, 151–161.
SUPPORTING INFORMATION
Supplementary data are available at GJI online.
Figure S1. Downsampling of the Sentinel-1 ascending track 116
LOS coseismic displacements using a resolution-based downsam-
pling scheme: (a) input full unwrapping interferogram; (b) resam-
pled interferogram. Squares indicate areas within which the LOS
displacements are averaged. (c) LOS profile across sections A–A’
showing the full unwrapped interferogram (coloured circles) versus
the resampled one (open circles). The pink star is the main shock
epicentre provided by AFAD.
Figure S2. Downsampling of the Sentinel-1 ascending track 043
LOS coseismic displacements using a resolution-based downsam-
pling scheme: (a) input full unwrapping interferogram; (b) resam-
pled interferogram. Squares indicate areas within which the LOS
displacements are averaged. (c) LOS profile across sections A–A’
showing the full unwrapped interferogram (coloured circles) versus
the resampled one (open circles). The pink star is the main shock
epicentre provided by AFAD.
Figure S3. Downsampling of the Sentinel-1 descending track 123
LOS coseismic displacements using a resolution-based downsam-
pling scheme: (a) input full unwrapping interferogram; (b) resam-
pled interferogram. Squares indicate areas within which the LOS
displacements are averaged. (c) LOS profile across sections A–A’
showing the full unwrapped interferogram (coloured circles) versus
the resampled one (open circles). The pink star is the main shock
epicentre provided by AFAD.
Figure S4. Downsampling of the Sentinel-1 descending track 021
LOS coseismic displacements using a resolution-based downsam-
pling scheme: (a) input full unwrapping interferogram; (b) resam-
pled interferogram. Squares indicate areas within which the LOS
displacements are averaged. (c) LOS profile across sections A–A’
showing the full unwrapped interferogram (coloured circles) versus
the resampled one (open circles). The pink star is the main shock
epicentre provided by AFAD.
Figure S5. (a) Data, (b) model and (c) residuals sampled points
from the unwrapped ascending track 116 interferogram. The violet
box represents our best-fitting uniform-slip solution. The pink star
is the main shock epicentre provided by AFAD.
Figure S6. (a) Data, (b) model and (c) residuals sampled points
from the unwrapped ascending track 043 interferogram. The violet
box represents our best-fitting uniform-slip solution. The pink star
is the main shock epicentre provided by AFAD.
Figure S7. (a) Data, (b) model and (c) residuals sampled points from
the unwrapped descending track 123 interferogram. The violet box
represents our best-fitting uniform-slip solution. The pink star is the
main shock epicentre provided by AFAD.
Figure S8. (a) Data, (b) model and (c) residuals sampled points from
the unwrapped descending track 021 interferogram. The violet box
represents our best-fitting uniform-slip solution. The pink star is the
main shock epicentre provided by AFAD.
Figure S9. Trade-off curve between solution roughness and
weighted misfit. The red circle indicates the chosen scalar smooth-
ing factor.
Figure S10. Relationship between RMS data reduction and relative
weighting factors. (a) The blue circles represent the RMS reduction
of the descending track 123, while the red ones are relevant to the
ascending track 116. (b) The green circles represent RMS reduction
of the ascending track 043 relative to both the descending track 123
(blue circles) and the ascending track 116 (red circles). (c) The light
blue circles show the RMS reduction of the descending track 021
relative to all the other data sets.
Figure S11. (a) Data, (b) model and (c) residuals sampled points
from the unwrapped ascending track 043 interferogram. The pink
star is the main shock epicentre provided by AFAD.
Figure S12. (a) Data, (b) model and (c) residuals sampled points
from the unwrapped descending track 021 interferogram. The pink
star is the main shock epicentre provided by AFAD.
Figure S13. Coseismic slip distributions on the causative fault seg-
ment from the inversion of (a) only the ascending tracks, (b) only
the descending tracks, and (c) the full data sets. The pink star is the
main shock epicentre provided by AFAD.
Please note: Oxford University Press is not responsible for the con-
tent or functionality of any supporting materials supplied by the
authors. Any queries (other than missing material) should be di-
rected to the corresponding author for the paper.
Downloaded from https://academic.oup.com/gji/article/223/2/1054/5874258 by INGV user on 27 September 2020
