A 220,000-year-long continuous large earthquake record on a slow-slipping plate boundary

Abstract

Large earthquakes (magnitude ≥ 7.0) are rare, especially along slow-slipping plate boundaries. Lack of large earthquakes in the instrumental record enlarges uncertainty of the recurrence time; the recurrence of large earthquakes is generally determined by extrapolation according to a magnitude-frequency relation. We enhance the seismological catalog of the Dead Sea Fault Zone by including a 220,000-year-long continuous large earthquake record based on seismites from the Dead Sea center. We constrain seismic shaking intensities via computational fluid dynamics modeling and invert them for earthquake magnitude. Our analysis shows that the recurrence time of large earthquakes follows a power-law distribution, with a mean of 1400 ± 160 years. This mean recurrence is notable shorter than the previous estimate of 11,000 years for the past 40,000 years. Our unique record confirms a clustered earthquake recurrence pattern and a group-fault temporal clustering model, and reveals an unexpectedly high seismicity rate on a slow-slipping plate boundary.

Full text

Lu et al., Sci. Adv. 2020; 6 : eaba4170 27 November 2020
SCIENCE ADVANCES | RESEARCH ARTICLE
1 of 10
GEOLOGY
A 220,000-year-long continuous large
earthquake record on a slow-slipping plate boundary
Yin Lu1,2,3,4*†, Nadav Wetzler5, Nicolas Waldmann2, Amotz Agnon6,
Glenn P. Biasi7‡, Shmuel Marco1
Large earthquakes (magnitude ≥ 7.0) are rare, especially along slow-slipping plate boundaries. Lack of large
earthquakes in the instrumental record enlarges uncertainty of the recurrence time; the recurrence of large
earthquakes is generally determined by extrapolation according to a magnitude-frequency relation. We enhance
the seismological catalog of the Dead Sea Fault Zone by including a 220,000-year-long continuous large earth-
quake record based on seismites from the Dead Sea center. We constrain seismic shaking intensities via computa-
tional fluid dynamics modeling and invert them for earthquake magnitude. Our analysis shows that the recurrence
time of large earthquakes follows a power-law distribution, with a mean of 1400 ± 160 years. This mean recur-
rence is notable shorter than the previous estimate of 11,000 years for the past 40,000 years. Our unique record
confirms a clustered earthquake recurrence pattern and a group-fault temporal clustering model, and reveals an
unexpectedly high seismicity rate on a slow-slipping plate boundary.
INTRODUCTION
The understanding of earthquakes, in general, and seismic hazard,
in particular, relies on our knowledge of past seismic history. The
longer a time window that a record spans, the better that under-
standing can be. However, large earthquakes [moment magnitude
(Mw) ≥ 7.0] usually have recurrence intervals longer than the time
span of modern seismograph operation of about a century and in-
frequently occur on individual faults (1). Paleoseismological trenching
at suitable sites can extend the record for surface rupturing earth-
quakes to the past few thousand years. Subaqueous paleoseismology
exploits lacustrine and marine sediments to retrieve much longer
records of paleoseismic shaking. These long records can probe our
understanding of the physical behavior of fault systems in the wake
of large seismic events and are therefore essential for improving
seismic hazard assessment.
The earthquake recurrence pattern is a key issue regarding the
understanding of fault behavior and seismic hazard assessment.
Regular recurrence patterns exist in records of thousands of years
on geometrically simple and fast-slipping strike-slip faults, such as
the southern onshore section of the Alpine Fault in New Zealand
(2) and Wrightwood Section of the San Andreas Fault (3). Earth-
quake recurrence patterns of slow-slipping faults (<5mm year−1),
e.g., the Dead Sea Fault (4,5), are more difficult to determine be-
cause they usually have longer interseismic intervals. Long and
continuous paleoseismic records of several 104 years on these slow-
slipping faults will improve statistical analyses and hazard assessments
because the length of the record can compensate for the uncertainty
in event identification and dating (3). In addition, these paleoseismic
records with extremely long time spans will be more representative
of the complete fault history (6).
The Dead Sea Fault is a left lateral, >1000-km-long strike-slip
plate boundary separating the African and Arabian plates (Fig.1A).
Short-term slip-rate estimates from GPS measurements of 4.2 to
5.8mm year−1 along the central to the southern part of the fault
(7,8) are similar to the long-term geologic rate (9). The Dead Sea
Basin is the deepest and largest continental tectonic structure along
this plate boundary and has a width of 15 to 20km and a length of
~150km (Fig.1). During the Quaternary, a sequence of terminal
water bodies occupied the basin. The location of the Dead Sea Basin
makes the soft and water-saturated sediments that accumulate in
the basin excellent recorders of seismic shaking. Seilacher (10) and
El-Isa and Mustafa (11) first hypothesized that the asymmetric folds
of unlithified sediments exposed at the eastern Dead Sea margin are
earthquake-triggered deformations. Along the western Dead Sea
margin, layers of shattered folds in the form of intraclast breccia
layers are common (12). The juxtaposition of these layers against
syn-depositional faults (13) and the time correspondence of historical
and archaeological earthquakes during the past 3 thousand years
(ka) (14–16) indicate that these layers are subaqueous seismites.
These pioneering studies set the stage for our current subaqueous
paleoseismic research.
Heifetz etal. (17) proposed that earthquake-forced shear, lead-
ing to sediment turbulence called the Kelvin-Helmholtz instability,
is a plausible driver for folding and shattering in the soft sediments
in the Dead Sea Basin. The laminated Dead Sea sediments consist of
stably stratified water-saturated mud in which the density increases
with depth, and thus are not susceptible to Rayleigh-Taylor instability,
which requires inverted densities (17). Further analysis shows that
the geometry of folded layer textures in the Dead Sea sediments
obeys a power-law of −1.9, similar to Kelvin-Helmholtz turbulence
in other environments (18). The shear seismic energy provided by
earthquakes may exceed the gravitational potential energy, thus allow-
ing the laminated and stably stratified layers to move horizontally in
1Department of Geophysics, Tel Aviv University, Tel Aviv 6997801, Israel. 2Dr. Moses
Strauss Department of Marine Geosciences, Leon H. Charney School of Marine
Sciences, University of Haifa, Mount Carmel 3498838, Israel. 3Sedimentology and
Marine Paleoenvironmental Dynamics Group, Institute of Earth Sciences, Heidelberg
University, Heidelberg 69120, Germany. 4Geomorphology and Quaternary Geology
Research Group, University of Liege, Liege 4000, Belgium. 5Geological Survey of
Israel, 32 Yeshayahu Leibowitz Street, Jerusalem 9692100, Israel. 6The Neev Center
for Geoinfomatics, Institute of Earth Sciences, Hebrew University, Jerusalem 9190401,
Israel. 7Nevada Seismological Laboratory, University of Nevada Reno, Reno, NV
89557, USA.
*Corresponding author. Email: yinlusedimentology@yeah.net, yin.lu@uibk.ac.at
†Present address: Sedimentary Geology Group, Institute of Geology, University of
Innsbruck, Innsbruck 6020, Austria.
‡Present address: Earthquake Science Center, U.S. Geological Survey, Pasadena,
CA 91106, USA.
Copyright © 2020
The Authors, some
rights reserved;
exclusive licensee
American Association
for the Advancement
of Science. No claim to
original U.S. Government
Works. Distributed
under a Creative
Commons Attribution
NonCommercial
License 4.0 (CC BY-NC).
on January 24, 2021http://advances.sciencemag.org/Downloaded from

Lu et al., Sci. Adv. 2020; 6 : eaba4170 27 November 2020
SCIENCE ADVANCES | RESEARCH ARTICLE
2 of 10
the same direction but with different velocities, creating shear that
localizes at the layer interfaces (17,18). Previous numerical simula-
tions (17,18) confirm the “intraclast breccia layer” as the final stage
of soft-sediment deformation. Moreover, these simulations confirm
that the observed textures are a proxy for shaking intensity by asso-
ciating the stage of the deformation (in different thicknesses) with
minimum ground accelerations.
The interpretation of deformed sediment layers has been the pri-
mary means for recovering paleoseismic records of different magni-
tudes and time spans along the central Dead Sea Fault for up to 60 ka ago
(4,16,19). However, previous paleoseismic records along the cen-
tral Dead Sea Fault (4,5, 19–21) are either short or incomplete in
their record of moderate earthquakes (5.0 < Mw < 7.0), and the con-
straints of local intensities and magnitudes are poor. Here, we in-
vestigate earthquakes recorded in the sedimentary sequence of the
deep ICDP (International Continental Scientific Drilling Program)
Core 5017-1 from the Dead Sea depocenter (Fig.1B; Materials and
Methods). The previous dating constrains the age of the 457-m-long
composite core spanning from ~220 ka ago to the present (fig. S1
and table S1) (22,23). We use this record to (i) identify and measure
all folded and brecciated layers, (ii) constrain the shaking intensities
of individual events via computational fluid dynamics modeling,
and (iii) assess the recurrence pattern of large earthquakes and fault
behavior model during the past 220 ka.
RESULTS
Earthquake indicators and paleoevents
As an ultimate repository for mass wasting (24) with an average
sedimentation rate of 2mm year−1 (25), the Dead Sea depocenter
provides the most complete record of earthquake shaking along the
plate boundary. Alternating laminae of white aragonite and dark
detritus that characterize the sedimentary sequence of the ICDP Core
5017-1 serve as sensitive markers for identifying earthquake-induced
deformation (24,26). We identify in situ folded layers and intraclast
breccia layer as earthquake indicators (seismites), based on their re-
semblance to outcrop observations of seismites that are known to
be earthquake induced (4,12–16).
In situ folded layers
Similar to the structures preserved in the Dead Sea margin outcrops,
the folded layers in the drilling core from the Dead Sea depocenter
also appear as folded aragonite-detritus laminae in various forms of
(i) linear waves (Fig.2, AandB), (ii) asymmetric billows (Fig.2,
CtoF), and (iii) coherent vortices (Fig.2,GtoJ) (18). These deli-
cate aragonite laminae are well preserved and can be traced in the
strata, indicating that the layers are deformed in situ and have not
undergone any notable transportation. That is, notable transporta-
tion would disaggregate and destroy the delicate submillimeter-
thick aragonite laminae. Layer-parallel displacements characterize
these in situ folded layers. The ICDP Core 5017-1 is positioned in
the center of the Dead Sea abyssal plain. This makes improbable
postdepositional causes for layer-parallel shears such as sloping
substrates or downhill water flow above the sediments. In total, we
identify 367in situ folded layers in the ICDP Core 5017-1 (table S2).
Figure S2 shows more examples of in situ folded layers in the drilling.
Intraclast breccia layer
Similar to the structures preserved in the Dead Sea margin out-
crops, this type of layer from the Dead Sea depocenter consists of
mixed aragonite-detritus laminae fragments (Fig.2,KtoM). The in
situ deformation process of this type of layer is recognized by (i) the
lack of erosion processes at the base of the layer and (ii) some re-
maining parts of in situ folded layer can be observed directly in the
base of the intraclast breccia layer (Fig.2,KtoM, and fig. S3).
Together, these features provide evidence for the evolution of the
intraclast breccia texture under seismic shaking and differentiate it
from any other detrital layers with laminae fragments that formed
by secondary sedimentary processes such as mass transport. In
total, we identify 46 intraclast breccia layers in the ICDP Core 5017-1
(table S2). Figure S3 shows more examples of intraclast breccia layers
in the core.
Our observations reveal that some in situ deformed layers have
been re-deformed by a latter deformation within a short core sec-
tion. Using geophysical and chemical datasets and careful sedimen-
tary structure analysis, we recognize these redeformed seismites.
Fig. 1. Tectonic setting of the Dead Sea Fault. (A) Dead Sea Fault is a sinistral
boundary between the African and Arabian plates (43). (B) Major active faults (43, 44)
along the plate boundary, Dead Sea Transform; in this area, the fault is composed
of four fault segments. The red star marks the drilling site; the black points mark
places referred to in the study; the magenta triangles indicate historic and instru-
mental Mw ≥ 6.0 earthquakes near the drilling site (45). (C) The gray bars represent
the fault rupture of historic Mw ≥ 7.0 earthquakes since 31 BCE (34, 36) that oc-
curred along the focused part of the fault—the central Dead Sea Fault (up to
150 km north and south of the drilling site).
on January 24, 2021http://advances.sciencemag.org/Downloaded from

Lu et al., Sci. Adv. 2020; 6 : eaba4170 27 November 2020
SCIENCE ADVANCES | RESEARCH ARTICLE
3 of 10
The potential uncertainties in thickness measurements (for rede-
formed seismites) that are induced by the redeformation range from
a few millimeters to several centimeters and therefore have no nota-
ble effects on shaking intensity estimation. In addition, regarding
the latter in situ deformations, the shape and thickness of total folded
sediments constrain the intensity of seismic shaking regardless of
the type and thickness of the redeformed seismites. In total, we
identify 413 independent seismic shaking markers from the Dead
Sea center.
Ground acceleration constraint for deformed layers
We update the previous computational fluid dynamics modeling of
Wetzler etal. (18) by extending the upper limit of layer thickness
and ground acceleration from 0.5m and 0.6g to 1.0m and 1.0g, re-
spectively (Fig.3; Materials and Methods). We run a series of two-
dimensional numerical simulations using the physical properties of
the soft sediments at the bottom of the Dead Sea based on the Kelvin-
Helmholtz instability mechanism. According to the numerical sim-
ulations, formation of a layer of (i) linear waves, (ii) asymmetric
billows, (iii) coherent vortices, and (iv) intraclast breccia requires a
minimum acceleration of 0.13, 0.18, 0.34, and 0.50g, respectively
(Fig.3). Also, our numerical simulations indicate that the thickness
of the deformed layer scales with acceleration (Fig.3 and fig. S4).
Because we could not directly determine the Reynolds number of
an individual unstable layer at the time of seismic shaking, we con-
strain only the lower boundary of acceleration needed to initiate
deformation of a layer with a certain thickness. By considering both
the shape and thickness of the deformed layers, we identified 18, 67,
139, 141, 240, and 413 events with acceleration of ≥0.65g, ≥0.50g,
≥0.34g, ≥0.26g, ≥0.18g, and ≥0.13g, respectively (table S2).
Return time statistics
We calculate the timing of the seismites between dated horizons by
linear interpolation between dated points of the cored section. Previous
computational fluid dynamics modeling indicates the sediment tur-
bulence develops at the interface of water and water-saturated sedi-
ments, and one earthquake produces only one in situ deformation
(17,18). Therefore, the age of the first undisturbed lamina overlying
the deformed layer constrains the time of formation of each seismite.
The 413 acceleration ≥0.13g events that Core 5017-1 records have a
mean return time of 530 ± 40 (SEM) years and an SD of 900 years
(Fig.4A and Table1). The strong seismic shaking events (acceleration,
≥0.34g) that Core 5017-1 records have a much longer mean return
time of 1500 ± 190 years and an SD of 2200 years. The acceleration
≥0.13, ≥0.18, ≥0.26, ≥0.34, ≥0.50, and ≥0.65g events have the
coefficient of variation (COV; SD divided by mean) of 1.6, 1.7, 1.5,
1.5, 1.6, and 1.5, respectively. Earthquake distributions with COV
≤ 0.7 are “quasi-periodic,” distributions with COV around 1 are
random, while distributions with COV > 1 are “clustered” (aperiodic)
(2,27). Thus, the events that Core 5017-1 records at different
shaking intensity levels are clustered in time, which is in line with
some previous paleoseismic investigations in the region (4,5). Be-
sides, the acceleration ≥0.13g and ≥0.34g events present a power-
law–like probability density of the recurrence interval (Fig.4,B
andC, and Table1).
DISCUSSION
Lower-bound magnitude determination for the paleoevents
Because of the intrinsic geometrical spreading, dissipation, and
possibly dispersion of the seismic energy, ground motion effects of
a moderate earthquake nearby generate similar shaking intensities
to those generated by a large earthquake farther away. Therefore,
determining the location of an earthquake along the length of the
Dead Sea Fault or on other nearby faults is the key for the magni-
tude constraint. It follows that magnitude estimation based on a
single station is difficult and dependent on the location of possible
source faults. For this sinistral boundary between the African and
Arabian plates, we model the potential source region of earthquakes
Fig. 2. Paleoearthquake indicators in the ICDP Core 5017-1. (A to J) In situ fold-
ed layers; (A and B) linear waves, (C to F) asymmetric billows, and (G to J) coherent
vortices. (K to M) Intraclast breccia layers. The vertical light blue bars indicate the position
of events. Core depth: (A) 11,010.0 to 11,012.0 cm; (B) 16,604.0 to 16,608.0 cm; (C) 10,929.9
to 10,932.4 cm; (D) 26,582.7 to 26,585.2 cm; (E) 32,861.0 to 32,862.5 cm; (F) 35,921.8
to 35,923.8 cm; (G) 13,754.4 to 13,758.0 cm; (H) 10,605.4 to 10,606.9 cm; (I) 36,425.9 to
36,427.9 cm; (J) 12,528.0 to 12,532.0 cm; (K) 14,492.5 to 14,500.0 cm; (L) 39,206.4 to
39,210.4 cm; and (M) 10,772.0 to 10,787.0 cm.
on January 24, 2021http://advances.sciencemag.org/Downloaded from

Lu et al., Sci. Adv. 2020; 6 : eaba4170 27 November 2020
SCIENCE ADVANCES | RESEARCH ARTICLE
4 of 10
as having a fixed width and a length of a few hundred kilometers (Fig.1).
The nearest faults are ~5km from the Core 5017-1 drilling site.
We apply three empirical attenuation relations (28–31) developed
for the Dead Sea region to constrain the lower magnitude limit of
paleoseismic events that the ICDP Core 5017-1 records (Materials
and Methods). Among them, two attenuation relations (28–30) are
described by macroseismic intensity, magnitude, and epicentral
distance (D), and one relation (31) is described by peak ground ac-
celeration (PGA), magnitude, and epicentral distance. To apply the
two attenuation relations that are described by seismic intensity, we
convert accelerations into seismic intensity via the linear relation-
ships between PGA and the modified Mercalli intensity scale (MMI)
that Wald etal. (32) proposed (table S2). According to the three
regional empirical attenuation relations, PGA ≥ 0.13g or MMI ≥
VI½ corresponds to Mw ≥ 5.5, 5.3, and 5.7, by taking Dmin = 5km
(table S3). Therefore, we interpret the lower-bound magnitude of
the recorded PGA ≥ 0.13g (MMI ≥ VI½) events as Mw ≥ 5.3. Con-
sidering that the incompleteness of moderate earthquakes in the
paleoseismic record, we infer that the mean recurrence of Mw ≥ 5.3
earthquake on the central Dead Sea Fault Zone is <530 ± 40 years.
This value is much shorter than the previously obtained mean re-
currence of ~1600 years for the same magnitude, based on the pale-
oseismic record between 60 and 14 ka ago (4).
Magnitude constraint for strong seismic shaking events
The ICDP Core 5017-1 records 139 strong seismic shaking events
with PGA ≥ 0.34g (MMI ≥ VIII) (table S2). According to the re-
gional empirical attenuation relations (28–31), one felt intensity, for
example, MMI ≥ VIII (PGA ≥ 0.34g) at the drill site, will require a
moderate earthquake (6.0 < Mw < 7.0) with a D between 5 and
30 km, an Mw ≥ 7.0 earthquake with a D of 30 km, or an Mw ≥ 8.0
earthquake with a D of up to 150km. We adopt a maximum magni-
tude in the region of Mw 8.0, which will require a rupture of ~300km
along a major fault. Therefore, we consider only large earthquakes
with D ≤ 150km (the central Dead Sea Fault Zone) as triggers of the
strong seismic shaking events (MMI ≥ VIII) that Core 5017-1 re-
cords. On this slow-slipping plate boundary, the Dead Sea Fault (zone)
is the only real contributing fault because most of the transform margin
slip rate is on the Dead Sea Fault, and the other faults were either far
away or had low slip rates. Within the Dead Sea region, no other
major faults have sufficient length and slip rate to materially con-
tribute to the rate of Mw ≥ 7 earthquakes. The historic earthquake
catalog from the region shows that the 1822, 1712, 1408, 1170,
1139/1140, and 859/860 CE Mw ≥ 7.0 earthquakes occurred with
distance to the drill site ≥300km north of the Dead Sea (the north-
ern Dead Sea Fault Zone). However, none of them have an expres-
sion in the Dead Sea Core 5017-1 record.
The spatial distribution of instrumental and historic moderate
and large earthquakes on the central Dead Sea Fault Zone during
the past 2 ka do supply additional clues for magnitude constraint
for these strong seismic shaking events. The instrumental (33) and
historical (5,34–36) earthquake catalogs reveal that during the past
2 ka, all major earthquakes (Mw ≥ 6.0) occurred with D ≥ 30km
from the drilling site (Fig.1B). By taking the past 2 ka earthquake
scenario as an analogy for the paleoseismic record, we assume that
most Mw ≥ 6.0 earthquakes occurred with D ≥ 30km from the
drilling site. Under this basic assumption and the three regional
empirical attenuation relations, (i) an intensity of MMI ≥ VIII
(PGA ≥ 0.34g) requires an earthquake with Mw ≥7.0, ≥7.0, and
≥7.3; (ii) an intensity of MMI ≥ VIII½ (PGA ≥ 0.50g) requires an
earthquake with Mw ≥7.4, ≥7.3, and ≥7.6; and (iii) an intensity of
MMI ≥ IX (PGA ≥ 0.65g) requires an earthquake with Mw ≥ 7.8,
≥7.6, and ≥7.8 (Fig.4E and table S3). Therefore, we interpret the
corresponding lower-bound magnitudes of strong seismic shaking
events in the ICDP Core 5017-1 to be Mw ≥ 7.0, 7.3, and 7.6, respectively.
We test our magnitude conversion versus known historic earth-
quakes on the central Dead Sea Fault Zone. Six seismites in Core
5017-1 dated at −2 ± 44 years before the present (yr B.P.), 42 ± 44 yr
Fig. 3. Numerical simulation on in situ folded layer and intraclast breccia structures in the Dead Sea sedimentary sequences. (A) Typical structures from Dead
Sea depocenter Core 5017-1. (B) Typical structures from Dead Sea onshore outcrops (Fig. 1B). (C) Schematic diagrams based on snapshots from the numerical simula-
tions demonstrating the four structures. (D) Quantitative estimation of the accelerations that are needed to initiate the four structures with different thicknesses; the
deformations normally occurred when Richardson number ≤ 0.125.
on January 24, 2021http://advances.sciencemag.org/Downloaded from

Lu et al., Sci. Adv. 2020; 6 : eaba4170 27 November 2020
SCIENCE ADVANCES | RESEARCH ARTICLE
5 of 10
B.P., 148 ± 44 yr B.P., 1248 ± 44 yr B.P., 1555 ± 47 yr B.P., and 1626 ±
47 yr B.P. correspond to the 1956 CE (Mw 5.5; D, ~5 km), 1927 CE
(Mw 6.25; D, ~30 km), 1834 CE (Mw ~ 6; D, ~60 km), middle
8th century (Mw > 7; D, ~100 km), 419 CE (Mw ~ 6; D, ~40 km), and
363 CE (Mw ~ 6.8; D, ~70 km) earthquakes, respectively (table S4)
(36). According to the regional empirical attenuation relations, we
constrain the magnitudes of the six paleoearthquakes (seismites) with
intensities of VII (0.18g), VI½ (0.13g), VI (0.09g), VII (0.18g), VI
(0.09g), and VII (0.18g) as Mw 5.6, Mw 6.1, Mw 6.2, Mw 7.1, Mw 6.0,
and Mw 6.9, respectively, which are in line with recorded historic
magnitudes (table S4). This test supports our magnitude conversion
based on the regional empirical ground motion attenuation relations.
Integrate a 220-ka-long large earthquake record
A few short gaps may exist in the paleoseismic record based on the
ICDP Core 5017-1, due to low coring recovery rate and halite depo-
sition at some depth. In addition, there is the possibility that some
remote Mw ~ 8.0 earthquakes with D > 150km induced the PGA <
0.34g (MMI < VIII) events. We incorporate previously established
large earthquake records from the central Dead Sea Fault Zone into
Fig. 4. Return time statistics of seismites and magnitude constraint for strong seismic shaking events during the past 220 ka. (A) Temporal distribution of moder-
ate (PGA ≥ 0.13g) and strong (PGA ≥ 0.34g) seismic shaking events. (B and C) Histograms for return times of PGA ≥ 0.13g and PGA ≥ 0.34g events. We plot two distribution
types (exponential and power-law) for each dataset. (D) Normalized return time data to show return time distribution of moderate and strong seismic shaking events.
(E) Magnitude constraint for strong seismic shaking events by applying the three regional empirical attenuation relations (28–31), taking the past 2-ka earthquake scenario
as an analogy for the paleoseismic record, and assuming that most Mw ≥ 6.0 earthquakes occurred with D ≥ 30 km from the drilling site (see the text for details); D, epicentral
distance. (F) Comparison of different temporal distributions of large earthquakes on the central Dead Sea Fault Zone derived from three different geological records.
(G) Magnitude-frequency distribution of modern (33) (olive colored) and paleoearthquakes (pink colored) on the central Dead Sea Fault during the past 220 ka; the number
of modern earthquakes (fig. S5) is extrapolated to 220 ka.
on January 24, 2021http://advances.sciencemag.org/Downloaded from

Lu et al., Sci. Adv. 2020; 6 : eaba4170 27 November 2020
SCIENCE ADVANCES | RESEARCH ARTICLE
6 of 10
our present 220-ka paleoseismic series to develop two constraining
limitations. Begin etal. (19) developed a 60-ka-long Mw ≥ 7.0 paleo-
seismic record based on seismites identified from Dead Sea western
onshore outcrops (Peratzim Valley; Figs.1B and 4F). In addition,
Kagan etal. (20) and Kagan (21) developed a 185-ka-long Mw ~8.0
paleoseismic record based on damaged cave deposits in the Soreq
and Har-Tuv Caves, located 40km due west of the Dead Sea Fault
(Figs.1B and 4F). The long record comprises 26 events that occurred
in the past ~185 ka.
Because the paleoseismic record of Begin etal. (19) is too short,
we incorporate the paleoseismic record of Kagan etal. (21) into
our present 220-ka paleoseismic series. Among the 26 events, four
speleoseismites correspond to PGA < 0.34g (MMI < VIII) seismites
(in situ deformed layers) in Core 5017-1, and four speleoseismites
have no corresponding seismites in the drilling (table S5). We,
therefore, incorporate these eight speleoseismites into the 220-ka
paleoseismic series. In addition, we incorporate four Mw ≥ 7.0 his-
toric earthquakes that occurred near the drilling site (D < 150 km)
during the past 2 ka (Fig.1C) into the 220-ka paleoseismic series,
because these four historic earthquakes either correspond to PGA <
0.34g (MMI < VIII) events or do not show discernible responses in
Core 5017-1 (table S5). In total, we incorporate 12 Mw ≥ 7.0 events
(including 8 Mw ~ 8.0 events) into the 220-ka paleoseismic series
(Fig.4F). Thus, the integrated 220-ka-long large earthquake record
comprises 151 Mw ≥ 7.0 events, 75 Mw ≥ 7.3 events, and 26 MMI ≥
IX (PGA ≥ 0.65g) events. The integrated record yields a mean re-
currence of 6200 ± 1800 years for the 26 largest events, similar to
Kagan etal.’s (21) mean recurrence of 6900 ± 1000 years for Mw
~8 earthquakes derived from the speleoseismite record. We there-
fore shift the magnitude corresponding to MMI ≥ IX (PGA ≥ 0.65g)
events from Mw ≥ 7.6 to Mw ≥ 7.8 (table S3), which is also in line
with the magnitudes converted from the other two regional empirical
attenuation relations (28,31).
A previous compilation of instrumental, historic, and paleoseismic
studies inferred that the Gutenberg-Richter distribution has been
stable in the Dead Sea region during the past 60 ka, with a b value of
~0.95 (37), and similar to a b value of 0.97 deduced from instru-
mental dataset (Fig.4G). The integrated 220-ka-long large earth-
quake record shows a b value of 0.95 (Fig.4G, magenta color),
assuming a Gutenberg-Richter distribution for the Dead Sea region
as a whole. This similar b value of the Gutenberg-Richter distribu-
tion and approximate match in a value after matching the catalog
time span to the 220-ka total length implies that our magnitude
constraint for strong seismic shaking events is appropriate, and our
record of large earthquakes is relatively complete.
Earthquake recurrence pattern and fault behavior model
and their implications for seismic hazard
The paleoseismic record of Begin etal. (19) yields a mean recurrence
of 4600 ± 1500 years for Mw ≥ 7.0 earthquakes and a COV of 1.0
during the past 60 ka (Fig.4F and Table 1). The exponential and
power- law distributions, however, are not fit well to the distribution
of these recurrence times (Fig.5A). In contrast to the paleoseismic
record of Begin etal. (19), the paleoseismic record of Kagan et al.
(21) yields a mean recurrence of 6900 ± 1000 years for Mw ~8.0
earthquakes and a COV of 0.7 (Fig.4F and Table1). The earthquake
recurrence times approximately follow a Weibull distribution (Fig.5B).
This type of distribution, together with the small COV, suggests a quasi-
periodic earthquake recurrence pattern for the largest earthquakes.
The integrated 220-ka-long Mw ≥ 7.0 earthquake record yields a
mean recurrence of ≤ 1400 ± 160 years and a COV of 1.4 (Fig.4F and
Table1). Recurrence times follow a power-law distribution (Fig.5C).
In contrast to the recurrence patterns revealed by the abovemen-
tioned two large earthquake records, the integrated 220-ka-long
Mw ≥ 7.0 record shows a clustered recurrence pattern (Fig.5D and
Table1). The distribution of earthquake recurrence times shows a
maximum probability density at the shortest recurrence times (< 0.7 ka)
and exhibits a long tail at the longest recurrence times (> 5.5 ka).
The high probability density at the shortest recurrence times may rep-
resent intra-cluster recurrence, and the long tail indicates intercluster
Table 1. Statistical analysis of recurrence times for the referred records.
Paleoseismic
record of (19)
Paleoseismic
record of (21)
PGA ≥ 0.13g (MMI
≥ VI½) events (this
study)
PGA ≥ 0.34g (MMI
≥ VIII) events (this
study)
Integrated Mw ≥
7.0 record (this
study)
Lower-bound magnitude (Mmin) 7.0 8.0 5.3 7.0 7.0
Time period Past 60 ka Past 185 ka Past 220 ka Past 220 ka Past 220 ka
Number of events 13 26 413 139 151
SD (year) 4800 4600 900 2200 2000
SEM (year) 1500 1000 40 190 160
Mean return time (year) 4600 ± 1500 6900 ± 1000 530 ± 40 1500 ± 190 1400 ± 160
COV 1.05 0.67 1.65 1.49 1.43
Fitting of return time
distribution (R2)
Weibull – 0.97 – – –
Exponential 0.01 0.85 0.72 0.63 0.66
Power-law 0.14 0.59 0.93 0.88 0.89
The best fit to the distribution – (Fig. 5A) Weibull (Fig. 5B);
y = 0.18x0.7e(−0.03*x^1.7)
Power-law (Fig. 4B);
y = 0.03x−1.5
Power-law (Fig. 4C);
y = 0.14x−1.2
Power-law (Fig. 5C);
y = 0.15x−1.3
Earthquake recurrence pattern Random Quasi-periodic Clustered
Fault behavior model – – The group-fault temporal clustering model
on January 24, 2021http://advances.sciencemag.org/Downloaded from

Lu et al., Sci. Adv. 2020; 6 : eaba4170 27 November 2020
SCIENCE ADVANCES | RESEARCH ARTICLE
7 of 10
recurrence in the group-fault temporal clustering model of McCalpin
(6). The group of faults in this case may be the bounding faults of
the Dead Sea Basin, plus ruptures within the basin itself. For example,
ruptures entering the basin from the south would cause strong shaking
at the drilling site and cluster by promoting rupture to the north on
the west-bounding fault. Additional Mw ≥ 6 events could be required
in the overlap zone to maintain the net plate boundary slip rate.
Lower apparent rates of large events outside the basin may correspond
to the longer intercluster times of the ICDP Core 5017-1 record.
The specific locations of the three study sites (Peratzim Valley,
Soreq and Har-Tuv Caves, and ICDP Core 5017-1; Fig.1B) could
partially explain the different earthquake recurrence patterns re-
vealed by the three large earthquake records. The Peratzim Valley is
situated west of the southern Dead Sea Basin. The seismites preserved
at the site may primarily record earthquakes originating from the
southernmost end of the central Dead Sea Fault and thus lead to the
aperiodic recurrence pattern with a much longer mean recurrence.
The Soreq and Har-Tuv Caves located on the northwestern side of
the Dead Sea Basin tend to be more frequently affected by the
movement of the northern part of the central Dead Sea Fault. This
location may not record some Mw ~ 8.0 earthquakes, which extend
south from the southern Dead Sea Basin. The speleoseismite record,
therefore, may correspond more closely with nearby faults and in a
quasi-periodic recurrence pattern. The ICDP Core 5017-1, with a
central location and much higher sediment accumulation rate, may
have recorded earthquakes from all faults that surround the drilling
site. The group-fault temporal clustering will appear if each quasi-
periodic behavioral fault is out of phase with each other. Alterna-
tively, the group-fault temporal clustering might arise from earth-
quake rupturing and stress transfer between different faults near
the drilling site over short time intervals and then followed by long
quiescence.
The unique location of the ICDP Core 5017-1 makes the inte-
grated 220-ka-long Mw ≥ 7.0 earthquake record the most complete
on the Dead Sea Fault and the most representative of the fault history.
The integrated record yields a mean recurrence of ≤ 1400 ± 160 years
for Mw ≥ 7.0 earthquakes, which is significantly shorter than the
mean recurrence of 4600 ± 1500 years for the same magnitude
based on Begin etal.’s 60-ka-long paleoseismic record (19). More-
over, our integrated record shows that Mw ≥ 7.0 earthquakes are
clustered during the past 40 ka (COV = 1.4), with a mean recur-
rence of ≤ 700 ± 110 years. Thus, the new results do not support the
previous conclusion that the seismic regime in the Dead Sea Basin
has been stationary for the past 40 ka and characterized by periodic
recurrence pattern with a mean recurrence of ~11 ka during that
time period (19). Therefore, our integrated 220-ka-long Mw ≥ 7.0
record indicates a higher seismic hazard than previously appreciat-
ed for the slow-slipping Dead Sea Fault.
Different from the periodic recurrence of earthquakes on fast-
slipping and geometrically simple strike-slip faults, e.g., the Alpine
Fault in New Zealand (2), we infer aperiodic earthquake behavior on
the slow-slipping and the geometrically complex sinistral boundary
between the African and Arabian plates. Our results suggest that re-
searchers may underestimate the seismic hazard potential of similar
Fig. 5. Recurrence pattern of large earthquakes (Mw ≥ 7.0) on the slow-slipping plate boundary during the past 220 ka. (A) Histograms for recurrence times
of Mw ≥ 7.0 events in the paleoseismic record of Begin et al. (19). (B) Histograms for recurrence times of Mw ~ 8.0 events in the paleoseismic record of Kagan et al. (21).
(C) Histograms for recurrence times in the integrated 220-ka-long Mw ≥ 7.0 record. (D) Normalized recurrence data for the three datasets for comparison.
on January 24, 2021http://advances.sciencemag.org/Downloaded from

Lu et al., Sci. Adv. 2020; 6 : eaba4170 27 November 2020
SCIENCE ADVANCES | RESEARCH ARTICLE
8 of 10
slow-slipping faults with irregular rupture. Our study highlights the
potential of in situ deformed sediment layers in a subaqueous environ-
ment as a strong-motion paleoseismometer to record long seismic
sequences covering multiple recurrence intervals of large earthquakes.
Long records are vital for accurate hazard assessment. Our quanti-
tative method of seismic record reconstruction, with paleoearthquake
intensity (ground acceleration) and magnitude estimation, may also
prove suitable for similar subaqueous environments along other faults.
MATERIALS AND METHODS
ICDP Core 5017-1 and age model
We retrieved the ICDP Core 5017-1 from the Dead Sea depocenter,
with a water depth of 297 m, from November 2010 to March 2011.
The 457-m-long composite core penetrated Holocene and late
Pleistocene sediments. Mud, alternations of aragonite-detritus laminae,
and halite comprised the sedimentary sequence. The millimeter- to
submillimeter-scaled detritus laminae are deposited during the rainy
seasons and are composed of allogenic quartz, calcite, and clay. The
delicate aragonite laminae are composed of authigenic aragonite
crystals (38). The core was dated with methods of 14C, U-Th, tuning,
and 18O stratigraphy correlation (22,23). Samples for 14C dating
focused on terrestrial plant remains to avoid a potential inheritance
effect from saline water. The measured 14C ages were calibrated by
using OxCal 4.3 and IntCal13 with 1 error. The U-Th dating was
carried out on primary aragonite laminae and reported with 2 error.
The age of Core 5017-1 ranges from ~220 ka ago to the present. In
total, we use 53 age points for the age model (fig. S1 and table S1).
We use linear interpolation between dated layers to calculate the
timing of each seismite. We did not consider it necessary to propa-
gate uncertainties of the age determination to each seismite. Previous
studies (2,3) have shown that when analytical uncertainties from
dating are small relative to the time between events, dating uncer-
tainty has little effect on parameter estimates such as COV and
virtually no effect on mean recurrence time.
Computational fluid dynamics modeling
Heifetz etal. (17) first modeled the physical process of soft-sediment
deformation preserved in the Dead Sea sediments. This linear fluid
dynamics modeling found that the Kelvin-Helmholtz instability is a
plausible mechanism for observed soft-sediment deformations.
Subsequently, Wetzler etal. (18) improved the modeling by consider-
ing nonlinear processes of soft-sediment deformation. Wetzler etal.
(18) examined more than 300 soft-deformation structures, which
include the linear waves, asymmetric billows, coherent vortices, and
intraclast breccia textures preserved in the late Quaternary Dead
Sea sedimentary sequences, and found that the geometry closely
follows a power-law (with an exponent of −1.9), compatible with
Kelvin- Helmholtz turbulence in other environments. The viscosity
and density of modern Dead Sea sediments in their water-saturated
condition are taken as a reasonable analog for late Quaternary Dead
Sea sediments and measured for modeling purposes. Boundary
conditions of the model were designed to reproduce different values
of ground acceleration. Minimum ground accelerations needed to
induce the layers of (i) linear waves, (ii) asymmetric billows, (iii)
coherent vortices, and (iv) intraclast breccia layers with different
thicknesses are determined.
In this study, we update the previous computational fluid dy-
namics modeling of Wetzler etal. (18) by extending the upper limit
of layer thickness and ground acceleration from 0.5m and 0.6g to
1.0m and 1.0g, respectively. We model the flow using the commercial
software Fluent (http://ansys.com/Products/Fluids/ANSYS-Fluent),
a computational fluid dynamics modeling tool. Wetzler etal. (18)
provide more information on the computational fluid dynamics
modeling.
The effects of water depth and secondary deformation were also
considered. The water depth itself does not directly affect the dy-
namics of Kelvin-Helmholtz instability between two mud layers in
the lakebed because they are in near hydrostatic balance before the
deformation. Thus, only the difference in densities between the layers
bears on the dynamics. The relative motion of the water layers with
respect to the mud delivers the energy via an internal gravity wave.
The dispersion relation for such waves is controlled by the thickness
of the deforming muddy layer h, up to an order of a meter. The
eigenfrequency (f) is
f = sqrt [ g(  m +  b ) / h(  m −  b ) ] (1)
where g is the effective acceleration of gravity, m is the density of
the minerals, and b is the density of the brine. With m/b of ~2
and brine content of around a third, f is approximately 1Hz.
Earthquake ground motion attenuation in the Dead Sea region
Three representative empirical ground motion attenuation relations
for the central Dead Sea Fault are available from the literature.
Intensity (I)–magnitude (M)–D relations
(i) Malkawi and Fahmi (28) developed empirical intensity (I)–magnitude
(M)–D relations characterizing earthquake ground motion attenu-
ation in Jordan and conterminous areas, based on instrumental and
historic datasets from the region. We chose to use their historic at-
tenuation relation instead of the instrumental relation because the
instrumental dataset does not include any large earthquakes, while
the historic dataset includes abundant large earthquakes. The equa-
tion is expressed as
I = 5.76 + 1.52 M S − 2.09ln(D + 25) (2)
When applying the relation, we convert the surface-wave magni-
tude (MS) scale into Mw scale according to (39,40).
(ii) Hough and Avni (29) published a large-magnitude attenua-
tion equation for the Dead Sea region calibrated using 133 seismic
intensity (I) determinations of the 1927 Mw 6.3 Jericho earthquake
I = − 0.64 + 1.7M − 0.004D − 1.67log(D) (3)
Darvasi and Agnon (30) improved the attenuation relation of
Hough and Avni (29) by considering local near-surface property
(shear wave velocity; Vs30) as an amplification factor. The calibrated
attenuation relation is expressed as
I = − 0.64 + 1.7M − 0.004D − 1.67log(D ) − 2.1ln( V s30 / 655) (4)
In this study, we apply this calibrated attenuation relation (Eq. 4)
instead of equation 3 of Hough and Avni (29). According to previous
shear wave measurements on the Dead Sea sediments (41), we use a
value of ~900 m/s for the Vs30 index. Following the attenuation rela-
tions (Eqs. 2 and 4), we estimate the change in M with D by fixing I.
on January 24, 2021http://advances.sciencemag.org/Downloaded from

Lu et al., Sci. Adv. 2020; 6 : eaba4170 27 November 2020
SCIENCE ADVANCES | RESEARCH ARTICLE
9 of 10
PGA-M-D relation
On the basis of 57 PGA recordings of 30 instrumental earthquakes
that occurred between 1979 and 2001 on the central and southern
Dead Sea Fault, Al-Qaryouti (31) proposed an attenuation relation
of PGA (g)
log(PGA ) = − 3.45 + 0.50M − 0.38log(D ) − 0.003D ± 0.31P (5)
in which P represents the normal distribution and P = 0 and 1 for
the 50th and 84th percentiles, respectively. Following the attenua-
tion relation, we estimate the change in M with D by fixing PGA.
These empirical attenuation relations provide the opportunity to
constrain magnitudes for our paleoseismic events recorded in the
ICDP Core 5017-1. However, we note that all three empirical atten-
uation relations have some shortcomings when applying them to our
220-ka paleoseismic record. In the attenuation relation of Malkawi
and Fahmi (28), 46 historic earthquakes were used to constrain the
magnitude scaling, far more than the other two attenuation rela-
tions. However, only 71 data points of maximum intensity were
included, and some magnitude uncertainties might exist in the his-
toric earthquakes (between 19 and 1896 CE). The attenuation rela-
tion of Hough and Avni (29) has robust intensity-distance scaling
and fit for the 1927 Mw 6.3 Jericho earthquake because 133 intensity
points were included. However, the attenuation relation was based
on only a single instrumental moderate earthquake. The attenuation
relation of Al-Qaryouti (31) is based on instrumental PGA values,
which provide high measurement accuracy. However, the number
of PGA points is limited because the observation time period is
relatively short (1979–2001). In addition, much of the data come
from single-site recordings. In the same year, Meirova etal. (42)
published an alternative PGA-M-D relation but limited to data
west of the Dead Sea fault. It predicted much higher magnitudes
than the relation of Al-Qaryouti (31) and was not applied to our
present study.
We applied all these three empirical attenuation relations (28,30,31)
to constrain the lower magnitude limit of paleoseismic events re-
corded in the ICDP Core 5017-1. The attenuation relation of Malkawi
and Fahmi (28) is more suitable for the inverting point intensity of
strong seismic shaking events (MMI ≥ IX) to magnitude than the
others. This is because the catalogs of Hough and Avni (29) and
Al-Qaryouti (31) used to reconstruct empirical attenuation relations
lack large earthquakes (Mw ≥ 7.0). We estimate that magnitude un-
certainties of historic earthquakes in the catalog of Malkawi and
Fahmi (28) do not have much impact on the strong seismic shaking
intensity range relevant to our present study.
b value and magnitude of completeness for the modern
earthquake catalog
We calculate the magnitude of completeness (Mc) for a rectangle of
an area 150km north and south to the drilling site, between 1990
and 2015 (fig. S5). Mc is computed iteratively from the goodness of
fit using a Kolmogorov-Smirnov test between observed and theo-
retical Gutenberg-Richter distributions with b value from maximum
likelihood estimation. We select the smallest magnitude for which
the difference between the distributions is negligible. More specifi-
cally, we use a threshold of D = 0.05 as the definition of negligible
when D is the Kolmogorov-Smirnov metric distance between the
two cumulative distribution functions, resulting in Mc = 3.1 and
b value = 0.97.
SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/
content/full/6/48/eaba4170/DC1
REFERENCES AND NOTES
1. R. J. Weldon, T. E. Fumal, G. P. Biasi, K. M. Scharer, Geophysics. Past and future
earthquakes on the San Andreas fault. Science 308, 966–967 (2005).
2. K. R. Berryman, U. A. Cochran, K. J. Clark, G. P. Biasi, R. M. Langridge, P. Villamor, Major
earthquakes occur regularly on an isolated plate boundary fault. Science 336, 1690–1693
(2012).
3. K. M. Scharer, G. P. Biasi, R. J. Weldon II, T. E. Fumal, Quasi-periodic recurrence of large
earthquakes on the southern San Andreas fault. Geology 38, 555–558 (2010).
4. S. Marco, M. Stein, A. Agnon, H. Ron, Long-term earthquake clustering: A 50,000-year
paleoseismic record in the Dead Sea Graben. J. Geophys. Res. 101, 6179–6191 (1996).
5. M. Ferry, M. Meghraoui, N. A. Karaki, M. Al-Taj, L. Khalil, Episodic behavior of the Jordan
Valley section of the Dead Sea Fault inferred from a 14-ka-long integrated catalog of large
earthquakes. Bull. Seismol. Soc. Am. 101, 39–67 (2011).
6. J. P. McCalpin, Application of paleoseismic data to seismic hazard assessment
and neotectonic research. Int. Geophys. 95, 1–106 (2009).
7. F. Masson, Y. Hamiel, A. Agnon, Y. Klinger, A. Deprez, Variable behavior of the Dead Sea
Fault along the southern Arava segment from GPS measurements. C. R. Geosci. 347,
161–169 (2015).
8. M. Sadeh, Y. Hamiel, A. Ziv, Y. Bock, P. Fang, S. Wdowinski, Crustal deformation along
the Dead Sea Transform and the Carmel Fault inferred from 12 years of GPS
measurements. J. Geophys. Res. 117, B08410 (2012).
9. S. Marco, Y. Klinger, Review of on-fault palaeoseismic studies along the Dead Sea Fault,
in Dead Sea Transform Fault System: Reviews, Z. Garfunkel, Z. Ben-Avraham, E. Kagan, Eds.
(Springer, 2014), pp. 183–205.
10. A. Seilacher, Sedimentary structures tentatively attributed to seismic events. Mar. Geol.
55, 1–12 (1984).
11. Z. H. El-Isa, H. Mustafa, Earthquake deformations in the Lisan deposits and seismotectonic
implications. Geophys. J. Int. 86, 413–424 (1986).
12. A. Agnon, C. Migowski, S. Marco, Intraclast breccias in laminated sequences reviewed:
Recorders of paleo-earthquakes. Geol. Soc. Am. Spec. Pap. 401, 195–214 (2006).
13. S. Marco, A. Agnon, Prehistoric earthquake deformations near Masada, Dead Sea graben.
Geology 23, 695–698 (1995).
14. Y. Enzel, G. Kadan, Y. Eyal, Holocene earthquakes inferred from a fan-delta sequence
in the Dead Sea graben. Quatern. Res. 53, 34–48 (2000).
15. R. Ken-Tor, A. Agnon, Y. Enzel, M. Stein, S. Marco, J. F. W. Negendank, High-resolution
geological record of historic earthquakes in the Dead Sea basin. J. Geophys. Res. 106 (B2),
2221–2234 (2001).
16. C. Migowski, A. Agnon, R. Bookman, J. F. W. Negendank, M. Stein, Recurrence pattern
of Holocene earthquakes along the Dead Sea transform revealed by varve-counting
and radiocarbon dating of lacustrine sediments. Earth Planet. Sci. Lett. 222, 301–314
(2004).
17. E. Heifetz, A. Agnon, S. Marco, Soft sediment deformation by Kelvin Helmholtz Instability:
A case from Dead Sea earthquakes. Earth Planet. Sci. Lett. 236, 497–504 (2005).
18. N. Wetzler, S. Marco, E. Heifetz, Quantitative analysis of seismogenic shear-induced
turbulence in lake sediments. Geology 38, 303–306 (2010).
19. Z. B. Begin, D. M. Steinberg, G. A. Ichinose, S. Marco, A 40,000 year unchanging seismic
regime in the Dead Sea rift. Geology 33, 257–260 (2005).
20. E. J. Kagan, A. Agnon, M. Bar-Matthews, A. Ayalon, Dating large infrequent earthquakes
by damaged cave deposits. Geology 33, 261–264 (2005).
21. E. Kagan, “Multisite late Quaternary paleoseismology in the Dead Sea transform region:
Independent recording by lake and cave sediments,” thesis, Hebrew University,
Jerusalem, Israel (2011).
22. H. Kitagawa, M. Stein, S. L. Goldstein, T. Nakamura, B. Lazar; DSDDP Scientific Party,
Radiocarbon chronology of the DSDDP core at the deepest floor of the Dead Sea.
Radiocarbon 59, 383–394 (2017).
23. S. L. Goldstein, Y. Kiro, A. Torfstein, H. Kitagawa, J. Tierney, M. Stein, Revised chronology
of the ICDP Dead Sea deep drill core relates drier-wetter-drier climate cycles to insolation
over the past 220 kyr. Quat. Sci. Rev. 244, 106460 (2020).
24. Y. Lu, N. Waldmann, G. I. Alsop, S. Marco, Interpreting soft sediment deformation
and mass transport deposits as seismites in the Dead Sea depocenter. J. Geophys. Res.
Solid Earth 122, 8305–8325 (2017).
25. Y. Lu, N. Waldmann, D. Nadel, S. Marco, Increased sedimentation following the Neolithic
Revolution in the Southern Levant. Global Planet. Change 152, 199–208 (2017).
26. Y. Lu, R. Bookman, N. Waldmann, S. Marco, A 45 kyr laminae record from the Dead Sea:
Implications for basin erosion and floods recurrence. Quat. Sci. Rev. 229, 106143 (2020).
27. Y. Y. Kagan, D. D. Jackson, Long-term earthquake clustering. Geophys. J. Int. 104, 117–133
(1991).
on January 24, 2021http://advances.sciencemag.org/Downloaded from

Lu et al., Sci. Adv. 2020; 6 : eaba4170 27 November 2020
SCIENCE ADVANCES | RESEARCH ARTICLE
10 of 10
28. A. I. H. Malkawi, K. J. Fahmi, Locally derived earthquake ground motion attenuation
relations for Jordan and conterminous areas. Q. J. Eng. Geol. Hydrogeol. 29, 309–319
(1996).
29. S. E. Hough, R. Avni, The 1170 and 1202 CE Dead Sea Rift earthquakes and long-term
magnitude distribution of the Dead Sea Fault zone. Isr. J. Earth Sci. 58, 295–308 (2009).
30. Y. Darvasi, A. Agnon, Calibrating a new attenuation curve for the Dead Sea region using
surface wave dispersion surveys in sites damaged by the 1927 Jericho earthquake.
Solid Earth 10, 379–390 (2019).
31. M. Y. Al-Qaryouti, Attenuation relations of peak ground acceleration and velocity
in the Southern Dead Sea Transform region. Arabian J. Geosci. 1, 111–117 (2008).
32. D. J. Wald, V. Quitoriano, T. H. Heaton, H. Kanamori, Relationships between peak ground
acceleration, peak ground velocity, and modified Mercalli intensity in California.
Earthq. Spectra 15, 557–564 (1999).
33. N. Wetzler, I. Kurzon, The earthquake activity of Israel: Revisiting 30 years of local
and regional seismic records along the Dead Sea Transform. Seismol. Res. Lett. 87, 47–58
(2016).
34. M. Lefevre, Y. Klinger, M. Al-Qaryouti, M. Le Béon, K. Moumani, Slip deficit and temporal
clustering along the Dead Sea fault from paleoseismological investigations. Sci. Rep. 8,
4511 (2018).
35. M. Meghraoui, Paleoseismic history of the Dead Sea Fault Zone. Encycl. Earthq. Eng.,
(2015).
36. A. Agnon, Pre-instrumental earthquakes along the Dead Sea rift, in Dead Sea Transform
Fault System: Reviews, Z. Garfunkel, Z. Ben-Avraham, E. Kagan, Eds. (Springer, 2014),
pp. 207–261.
37. Y. Hamiel, R. Amit, Z. B. Begin, S. Marco, O. Katz, A. Salamon, E. Zilberman, N. Porat, The
seismicity along the Dead Sea Fault during the last 60,000 years. Bull. Seismol. Soc. Am. 99,
2020–2026 (2009).
38. A. Haliva-Cohen, M. Stein, S. L. Goldstein, A. Sandler, A. Starinsky, Sources and transport
routes of fine detritus material to the Late Quaternary Dead Sea basin. Quat. Sci. Rev. 50,
55–70 (2012).
39. T. C. Hanks, D. M. Boore, Moment-magnitude relations in theory and practice. J. Geophys.
Res. Solid Earth 89, 6229–6235 (1984).
40. J. P. McCalpin, in Paleoseismology, J. P. McCalpin, Ed. (Elsevier, 2009).
41. J. B. Williams, “Estimation of earthquake source parameters from soft sediment
deformation layers present in Dead Sea muds” (Department of Civil Engineering,
California State University, 2004).
42. T. Meirova, R. Hofstetter, Z. Ben-Avraham, D. M. Steinberg, L. Malagnini, A. Akinci,
Weak-motion-based attenuation relationships for Israel. Geophys. J. Int. 175, 1127–1140
(2008).
43. Z. Ben-Avraham, Z. Garfunkel, M. Lazar, Geology and evolution of the southern Dead Sea
Fault with emphasis on subsurface structure. Annu. Rev. Earth Planet. Sci. 36, 357–387 (2008).
44. Y. Bartov, A. Agnon, Y. Enzel, M. Stein, Late Quaternary faulting and subsidence in the
central Dead Sea basin. Isr. J. Earth Sci. 55, 17–31 (2006).
45. T. Levi, D. Bausch, O. Katz, J. Rozelle, A. Salamon, Insights from Hazus loss estimations
in Israel for Dead Sea Transform earthquakes. Nat. Hazards 75, 365–388 (2014).
Acknowledgments: We appreciate the three anonymous reviewers for thorough,
constructive, and encouraging review and their strong recommendation. We thank J. Hall for
English editing. We also thank Z. Guo, Q. S. Liu, and Y. John Chen for supporting Y.L.’s 3-week
visit during September 2017 at the Department of Ocean Science and Engineering, Southern
University of Science and Technology, China, to continue working on the project. Funding:
This research was supported by the University of Liege under Special Funds for Research,
IPD-STEMA Program (R.DIVE.0899-J-F-G to Y.L. between 2019 and 2020), Post-Doctoral
Fellowship of the Faculty of Exact Sciences at Tel Aviv University (to Y.L. between 2016 and
2017), the Israel Science Foundation (Center of Excellence grant #1436/14 and grant #1645/19
to S.M.), the International Continental Scientific Drilling Program (ICDP), the DESERVE (Dead
Sea Research Venue) Virtual Institute under the auspices of the Helmholtz Association (https://
deserve-vi.net/) (to A.A.), and the Israel Science Foundation (ISF 363/20 to N.We.). Author
contributions: Y.L. put forward the path to the goal, developed the geological observations
and interpretations, and prepared the manuscript, which was edited by all the coauthors;
N.We. contributed to the numerical simulation; N.Wa. participated in the ICDP campaign and
the associated sampling and logging party together with A.A., who also contributed to the
interpretation of historic earthquake catalog and the inverting point intensity for magnitude-
frequency relations; G.P.B. contributed to the inverting point intensity for magnitude-
frequency relations and English editing; S.M. initiated and contributed to all aspects of the
project. Competing interests: The authors declare that they have no competing interests.
Data and materials availability: All data needed to evaluate the conclusions in the paper are
present in the paper and/or the Supplementary Materials. Correspondence and requests for
additional data should be addressed to Y.L. (yinlusedimentology@yeah.net; yin.lu@uibk.ac.at).
Submitted 2 December 2019
Accepted 16 October 2020
Published 27 November 2020
10.1126/sciadv.aba4170
Citation: Y. Lu, N. Wetzler, N. Waldmann, A. Agnon, G. P. Biasi, S. Marco, A 220,000-year-long
continuous large earthquake record on a slow-slipping plate boundary. Sci. Adv. 6, eaba4170
(2020).
on January 24, 2021http://advances.sciencemag.org/Downloaded from

A 220,000-year-long continuous large earthquake record on a slow-slipping plate boundary
Yin Lu, Nadav Wetzler, Nicolas Waldmann, Amotz Agnon, Glenn P. Biasi and Shmuel Marco
DOI: 10.1126/sciadv.aba4170
(48), eaba4170.6Sci Adv
ARTICLE TOOLS http://advances.sciencemag.org/content/6/48/eaba4170
MATERIALS
SUPPLEMENTARY http://advances.sciencemag.org/content/suppl/2020/11/19/6.48.eaba4170.DC1
REFERENCES http://advances.sciencemag.org/content/6/48/eaba4170#BIBL
This article cites 39 articles, 9 of which you can access for free
PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions
Terms of ServiceUse of this article is subject to the
is a registered trademark of AAAS.Science AdvancesYork Avenue NW, Washington, DC 20005. The title
(ISSN 2375-2548) is published by the American Association for the Advancement of Science, 1200 NewScience Advances
License 4.0 (CC BY-NC).
Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial
Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of
on January 24, 2021http://advances.sciencemag.org/Downloaded from