Content uploaded by Eleftheria Papadimitriou
Author content
All content in this area was uploaded by Eleftheria Papadimitriou
Content may be subject to copyright.
Seismicity Variations in the Southern Aegean, Greece, Before and After
the Large (M7.7) 1956 Amorgos Earthquake Due to Evolving Stress
ELEFTHERIA PAPADIMITRIOU,
1
GEORGIOS SOURLAS,
1
and VASSILIOS KARAKOSTAS
1
Abstract—The largest earthquake (M
0
¼4.910
27
dyncm) of the 20
th
century in the territory of Greece
occurred south of Amorgos Island, causing extensive destruction in the southern Aegean area. It occurred
on an ENE–trending normal fault that is seated parallel to the Island’s southern coastline. Changes in the
rates of moderate–size earthquakes (M 5.0) that occurred before and after the Amorgos earthquake,
within circular regions centered on its epicenter with radii of 100, 150 and 200 km, are investigated. The
rate for moderate–size events just before the main shock appears to be considerably increased when
compared to those of either preceding or subsequent periods. Further inspection reveals that more evident
seismicity fluctuations are attributed to distances exceeding 100 km. These changes may be indicative of a
broad region that is approaching a high stress state prior to an eventual large earthquake. Close to the
main event, that is, within the 100–km radius, a remarkable quiescence period lasting about two decades
before its occurrence was observed. Changes in seismicity are discussed in combination with static stress
changes calculated by the application of the stress evolutionary model that takes into account the coseismic
slip associated with the larger events (M 6.5) since the beginning of the 20
th
century and the tectonic
loading on the major faults in the study area. These larger events, as with the intermediate magnitude
seismicity taking place at distances exceeding 100 km and which encircled the quiescent area observed
during the last 22 years before the Amorgos earthquake, are well correlated with stress-enhanced areas in
each stage of the evolutionary model.
Key words: Seismicity rates, Coulomb stress changes, triggering, southern Aegean.
1. Introduction
The study area belongs to the extensional backarc Aegean region that includes
southern Bulgaria and former Yugoslavia, northern and central Greece, southern
Aegean volcanic arc, and southwestern and centralwestern Turkey (Fig. 1). The
Aegean region is one of the most active tectonic regions of the Alpine–Himalayan
belt, with its most prominent tectonic feature the subduction of the eastern
Mediterranean lithosphere under the Aegean Sea (PAPAZACHOS and COMNINAKIS,
1970) along the Hellenic arc. The seismicity is very high throughout the arc, which is
dominated by thrust faulting in a NE–SW direction of the axis of maximum
1
Geophysics Department, School of Geology, Aristotle University of Thessaloniki, GR54124
Thessaloniki, Greece. E-mail: ritsa@geo.auth.gr, vkarak@geo.auth.gr
Pure appl. geophys. 162 (2005) 783–804
0033 – 4553/05/050783 – 22
DOI 10.1007/s00024-004-2641-z
Birkha
¨user Verlag, Basel, 2005
Pure and Applied Geophysics
compression. A belt of thrust faulting runs along the southwestern coast of
Yugoslavia and continues south along the coastal regions of Albania and
northwestern Greece, resulting from continental collision between Outer Hellenides
and the Adriatic microplate. The direction of the maximum compression axis is
almost normal to the direction of the Adriatico–Ionian geological zone. Between
continental collision to the north and oceanic subduction to the south, the dextral
strike–slip Cephalonia Transform Fault (CTF) is observed (SCORDILIS et al., 1985) in
agreement with the known relative motion of the Aegean and eastern Mediterranean.
MCKENZIE (1978) showed that the northward motion of the Arabian plate pushes
the smaller Anatolian plate westward along the North Anatolian fault, continuing
along the North Aegean Trough (NAT) region, which is the boundary between the
Eurasian and south Aegean plates. Right–lateral strike–slip motion associated with
the North Anatolian Fault (NAF) appears to become more distributed in the North
Aegean Sea, which is characterized by a combination of right–lateral shear and
extension. This motion is transferred into the Aegean but in a southwesterly
direction.
Figure 1
Main seismotectonic properties of the Aegean and surrounding regions. The epicenter of the Amorgos
main shock is denoted by a star.
784 E. Papadimitriou et al. Pure appl. geophys.,
The Amorgos main shock is the largest event (M7.7) that occurred in the Aegean
region during the instrumental era. Since it is well accepted that significant variations
in seismicity behavior have been observed in association with the occurrence of large
earthquakes, the present study focuses on the identification of such patterns and their
interpretation. FEDOTOV (1965) first noticed that the seismic activity increased and
decreased within the seismic cycle. An increase in the level of seismicity in
surrounding regions, defining a quiescent region in the center and forming the
‘‘doughnut pattern’’ before great shallow earthquakes in Japan, was found and
demonstrated by MOGI (1969, 1981). This is attributed to the secondary seismic
activity associated with the many smaller faults that are located near the fault
associated with the largest shock. Since faults interact through their stress field, it is
expected that changes in stress will be reflected to the seismicity variations.
Increased interest has been raised among scientists concerning study of increased
seismicity prior to large earthquakes. ELLSWORTH et al. (1981) reported an increase
in the rate of earthquakes of M 5 in the years prior to the 1906 San Francisco
earthquake in a broad region covering much of the San Francisco Bay area. SYKES
and JAUME
´(1990) found that the rate of seismic moment release for events of
magnitude 5.0 M < 6.5 accelerated prior to the 1868, 1906, and 1989 earthquakes
in northern California and for shocks of 4.0 M < 6.0 before the southern
California event of M6.0 in 1948. KNOPOFF et al. (1996) found that all eleven
earthquakes in California with magnitudes greater than or equal to 6.8 from 1941 to
1993 were preceded by an increase in the rate of occurrence of earthquakes
registering magnitudes greater than 5.1. This precursory activity was concentrated in
regions with linear dimensions of the order of a few hundred kilometers, significantly
larger than the estimated fracture lengths of the ensuing strong earthquakes. The
pattern was almost unidentifiable for earthquakes with magnitudes less than about
4.6. BOWMAN et al. (1998) found that before all earthquakes of M 6.5 from 1950 to
1998 along the San Andreas system in California, there was a well–defined period of
accelerating seismic energy release within a finite region that scaled with the size of
impending large earthquakes. The rate of seismic moment release before large
earthquakes was studied for the area of Greece by PAPAZACHOS and PAPAZACHOS
(2000), who found that accelerating deformation has been observed for a few decades
before the occurrence of twenty–four strong (M ¼6.0–7.5) earthquakes. For the
Amorgos earthquake in particular, the accelerated deformation period started in
1920 and was associated with earthquakes of M 5.2 that occurred in an elliptical
area whose long axis was equal to 430 km.
The aim of this study is to examine the behavior of moderate–size (M ‡5.0)
seismicity before and after the 1956 Amorgos large (M7.7) earthquake in the
southern Aegean. The current analysis deals with the earthquake spatio–temporal
distributions as retroactively as the data permit, and the determination of the
observed patterns in relation with known physical models. Changes in seismicity
reflect changes in the stress field, and therefore the evolution of the stress field is
Vol. 162, 2005 Seismicity Variations in the Southern Aegean 785
examined. This has been done by calculating the Coulomb stress changes associated
with the occurrence of the larger (M ‡6.5) shocks, including in the calculations the
long–term tectonic loading on the major faults in the study area.
2. Changes in Frequency of Events before and after the Amorgos Earthquake
The occurrence rate of events above some magnitude in a given period is used as a
quantitative measure of the seismicity of a region. Seeking seismicity patterns before
and after the Amorgos main shock in an area surrounding its epicenter, the catalog
of earthquakes that occurred during the 20
th
century (PAPAZACHOS et al., 2000), that
is during the instrumental era, has been taken into account. The catalog is
homogeneous as all magnitudes are expressed as equivalent moment magnitudes
(PAPAZACHOS et al., 1997), and complete at magnitude 4.5 since 1950 and magnitude
5.0 since 1911. We are interested in examining how the seismic activity was affected
both temporarily and spatially and both before and after the occurrence of the
Amorgos main shock. Therefore, the events that occurred inside circular areas
centered on the main shock epicenter and having radii equal to 100, 150 and 200 km
are considered, in order to compare seismic activity within circular regions of varying
size. The choice of the radii is arbitrary albeit the first radius is comparable to the
main rupture length and the bigger one is comparable to the extent of the volume
affected by the preparation process of large earthquakes (e.g., PAPAZACHOS and
PAPAZACHOS, 2000). It resembles the analysis of DUand SYKES (2001) who
considered circular areas and equal–area annuli centered on the epicenter of the
Landers event and found the larger changes to occur about 150 km from it.
The problem of choosing the appropriate magnitude cutoff is solved by inspecting
the changes of occurrence rate of events with magnitude larger than or equal to 5.0.
From a visual inspection of the occurrence rate plots, it was found that the data were
complete for M > 5.0 only after 1920. Although we are interested in moderate
magnitude events (level of 2 to 3 orders of magnitude smaller than the main shock),
the changes for events of 4.5 £M£4.9 were also examined for the period of this
data sample’s completion. Even after the occurrence of the main shock no substantial
changes in the occurrence of smaller (M < 5.0) events appeared.
Figure 2 shows graphs of the cumulative number of events with varying
magnitude cutoffs that are plotted against time. It evidences that periods of increased
activity alternate with periods of relative quiescence. The definition of the time
intervals spanning each period is a reconciliation of all distances and magnitude
ranges. More specifically,
For events inside the circular area of R ¼100 km: An activation period started in
1920 and lasted until 1932 for events with 5.0 £M£5.4 (Fig. 2a) not observed for
5.5 £M£5.9 (Fig. 2b), culminated with the 1933 strong earthquake (M6.6), located
115 km from the Amorgos epicenter. A period of quiescence for all magnitude ranges
786 E. Papadimitriou et al. Pure appl. geophys.,
followed during 1933–1955, that is before the Amorgos main shock, the activity
heightened again during 1957–1983, while it decreased again during the last period,
1984–2002. The most prominent change for this circular area and for all the
magnitude ranges is the dramatic change in the immediate vicinity of the main shock,
Figure 2
(a) Cumulative number of earthquakes of 5.0 M5.4 that occurred during 1900–2002 inside a circular
area centered on the Amorgos main shock and with a radius R ¼100 km, as a function of time. (b) Same
as (a) but for events of 5.5 M5.9. (c) Same as (a) but for R ¼150 km. (d) Same as (b) but for
R¼150 km. (e) Same as (a) but for R ¼200 km. (f) Same as (b) but for R ¼200 km.
Vol. 162, 2005 Seismicity Variations in the Southern Aegean 787
of the total quiescence prior to an excitation afterwards, which is attributed to its
aftershock activity endured several years as is expected for large magnitude events.
For events inside circular areas of R ¼150 km and R ¼200 km: Four significant
changes in the rate of activity in a pattern converse to the previous one are observed
(Figs. 2c,d,e,f). The frequency of events increased after 1933 until 1956 with the
Amorgos main shock occurrence, becoming more spectacular for larger events
(Figs. 2d,f). This increase is interposed between two periods of relatively lower
occurrence rate in 1920–1933 and 1957–1983. A relative increase corresponds to the
period 1984–2002. These changes evidence two periods of high seismicity and two
periods of low seismicity in the study area. The pattern persists for both circular
areas and is more prominent for events with magnitude equal to or larger than 5.2
(Figs. 2c,e), which agrees with the finding of PAPAZACHOS and PAPAZACHOS (2000) as
mentioned earlier. The data inside the annuli of 100–150 km and 150–200 km are
now considered and their occurrence frequency is plotted in Figure 3. The excitation
period 1933–1955 is more evident for the annuli 100–150 km and for all magnitude
Figure 3
(a) Cumulative number of earthquakes of 5.0 M5.4 that occurred during 1900–2002 within annuli of
100–150 km, as a function of time. (b) Same as (a) but for events of 5.5 M5.9. (c) Same as (a) for
annuli of 150–200 km. (d) Same as (b) for annuli of 150–200 km.
788 E. Papadimitriou et al. Pure appl. geophys.,
ranges (Figs. 3a,b), as well as the relatively quiescent period (1957–1983) for both
annuli.
3. Spatial Distribution of Seismicity
In order to examine how the activity was distributed over the study area during
the four periods as defined in the previous section, the epicenters of events with
M‡5.0 were plotted in Figure 4. During 1920–1932, the epicenters occupy all the
study area, with the ones of larger magnitude events (5.5 £M£5.9) being inside the
annuli of 150–200 km from the main shock epicenter (Fig. 4a). During 1933–1955,
inside the area with a radius of 100 km from the main shock epicenter, a lack of
Figure 4
(a) Spatial distribution of earthquakes with M 5.0 that occurred during 1920–1933. Circles are centered
on the Amorgos epicenter and have radii equal to 100, 150 and 200 km. (b) Same as (a) for the period
1933–1955. (c) Same as (a) for the period 1957–1983. (d) Same as (a) for the period 1984–2002.
Vol. 162, 2005 Seismicity Variations in the Southern Aegean 789
activity is observed except one event in the magnitude range 5.0 £M£5.4, while the
number of larger events was dramatically increased in longer distances, most of them
being inside the annuli of 100–150 km (Fig. 4b). It is worthy to note that all the
instrumental events in our study area with M 6.5 occurred during this period.
During the period following the Amorgos main shock (1957–1983) the number of
smaller magnitude events increased with the larger magnitude events occurring closer
to the Amorgos epicenter (Fig. 4c). During the last period (1984–2002) the activity
was mainly concentrated in the eastern part of the study area and in longer distances
from the Amorgos epicenter than during the previous period (Fig. 4d).
The data during 1956 were not included in the plots of Figure 4, to avoid
contamination of the spatial pattern with the imminent foreshocks or aftershocks of
Amorgos event. Figure 5a shows the three events with 5.0 M5.4 that occurred
during the first six months of 1956 at distances of 184–198 km, two of them having
the same location. The seismic activity during July 9–December 31, 1956 constituted
two clearly distinguished clusters (Fig. 5b). The first cluster comprised the Amorgos
aftershocks, well confined along its fault, and the activation of the neighboring
Santorini fault (westward prolongation of the Amorgos fault) with an event of
M¼6.9 some minutes afterwards (Table 1). The second cluster defined the spatial
extent of a sequence that took place within a mean distance of 94 km south of the
Amorgos fault, with a main shock of M ¼6.0 associated with the Dionisades fault
on July 30 (Table 1). Although considerable uncertainties in epicentral determination
exist in comparison with the fault dimension, due to the station coverage at that time,
Figure 5
(a) Spatial distribution of earthquakes with M 5.0 that occurred during Jan. 1 – July 8, 1956. Circles are
centered on the Amorgos epicenter and have radii equal to 100, 150 and 200 km. (b) Same as (a) for the
period July 9 – Dec. 31, 1956.
790 E. Papadimitriou et al. Pure appl. geophys.,
Table 1
Information on the normal faults associated with the occurrence of known strong earthquakes in the study area (modified from PAPAZACHOS et al., 2001 and
PAPAZACHOS and PAPAZACHOU, 2002)
Fault Name Center Strike, deg. Dip, deg. Rake, deg. Length, km Slip rate
mm yr1
Occurrence
year
and magnitude
Latitude (N) Longitude (E)
1. Torbali 38.18 27.45 83 45 )115 50 4 1928, 6.5
2. Samos 37.71 26.85 91 45 )115 45 4 1904, 6.8
3. W. Buyuk 37.715 27.4 91 45 )115 45 4 1955, 6.9
4. Santorini 36.53 25.52 50 40 )90 40 4 1956, 6.9
5. Amorgos 36.73 25.99 65 40 )90 75 4 1956, 7.7
6. Kos 36.75 27.19 65 50 )90 55 4 1933, 6.6
7. Marmaris 37.03 28.11 80 42 )99 58 4 1869, 6.8
8. Simi 36.36 27.63 250 48 )78 35 4 1869, 6.8
9. Katavia 35.83 27.56 184 47 )98 20 3 1996, 6.2
10. Karpathos 35.76 27.05 185 47 )98 60 3 1948, 6.9
11. Zakros 35.16 26.49 14 47 )98 40 3 1922, 6.8
12. Dionisades 35.60 25.97 10 47 )98 20 3 1956, 6.0
13. Ierapetra 35.15 25.7 10 47 )98 65 3 1815, 6.8
Vol. 162, 2005 Seismicity Variations in the Southern Aegean 791
one can unambiguously observe that the epicenters are aligned in a N–S direction,
compatible with the E–W extension prevailing in the area (PAPAZACHOS et al., 1998).
4. Evolving Stress Field and Earthquake Occurrence
The specific seismicity behavior that was observed and described above could be
attributed to the changes in the stress field. These changes are associated with the
coseismic displacements of the largest events (M 6.5) and the long–term tectonic
loading on the major faults of the study area (DENG and SYKES, 1997). Interseismic
stress accumulation between large events is modeled by ‘‘virtual negative displace-
ments’’ along major faults in the entire region under study, using the best available
information on long–term slip rates. These virtual dislocations are imposed on the
faults with a sense of slip opposite to the observed slip. The magnitude of this virtual
slip is incremented according to the long–term rate of the fault. This virtual negative
slip is equivalent to constant positive slip extending from the bottom of the
seismogenic layer to infinite depth. Hence, tectonically induced stress builds up in the
vicinity of faults during the time intervals between earthquakes. All computed
interseismic stress accumulation is associated with the deformation caused by the
time–dependent virtual displacement on major faults extending to the depth at which
earthquakes and brittle behavior cease (15 km). Stress build-up is released entirely
or in part during the next large earthquake ruptures, which are considered positive in
the model. Changes in stress associated with large earthquakes are calculated for
coseismic displacements on each ruptured fault segment and by adding the changes in
the components of the stress tensor as they occur in time. Stress changes associated
with both the virtual dislocations and actual earthquake displacements are calculated
using a dislocation model (STEKETEE, 1958; OKADA, 1992; G. Converse, U. S.
Geological Survey, unpublished report, 1973).
Earthquakes occur when the stress exceeds the strength of the fault. The closeness
to failure was quantified by using the change in Coulomb failure function ðDCFF Þ
(modified from SCHOLZ, 1990). It depends on both changes in shear stress, Ds;and
normal stress, Dr:
DCFF ¼Dsþl0Dr:ð1Þ
Here l0is the apparent coefficient of friction. Both Dsand Drare calculated for
the fault plane of the next earthquake in the sequence of events whose triggering is
inspected. The change in shear stress, Dsis positive for increasing shear stress in the
direction of relative slip on the observing fault; Dris positive for increasing tensional
normal stress. When compressional normal stress on a fault plane decreases, the
static friction across the fault plane also decreases. Both positive Dsand Drmove a
fault toward failure; negative Dsand Drmove it away from failure. A positive value
of DCFF for a particular fault denotes movement of that fault toward failure (that is,
792 E. Papadimitriou et al. Pure appl. geophys.,
the likelihood that it will rupture in an earthquake is increased). The shear modulus
and Poisson’s ratio are fixed as 33 GPa and 0.25, respectively. The selection of the
value of the apparent coefficient of friction, l0, is based on previous results. A value
of l0equal to 0.4 was chosen and considered sufficient throughout the calculations
(NALBANT et al., 1998; KING et al., 1994a).
Since the study area is dominated by extension, the normal faults that are known
to be associated with events of M 6.0 since the 19
th
century (PAPAZACHOS et al.,
2001) are considered here (Fig. 6). On the basis of existing data, it is possible to
estimate slip rates for the faults of interest in the present study from the relative
motions between GPS stations straddling them. Such information is available from
MCCLUSKY et al. (2000) who interpreted GPS measurements of crustal motions for
the period 1988–1997. Based on the motion of specific GPS stations, the extension
velocities for each of these faults are defined approximately, so that their sum is in
accordance with the generally accepted motion of the south Aegean microplate
(Table 1). Nevertheless, more accurate long–term rates for each fault that contribute
Figure 6
Morphology and major faults of the study area (after PAPAZACHOS et al., 2001).
Vol. 162, 2005 Seismicity Variations in the Southern Aegean 793
to the total plate motion undoubtedly will contribute to better estimates of
earthquake hazard.
Fault length and average displacement are two parameters necessary for the
model application. We are interested in fault lengths expressing the main rupture as
well as the average value of coseismic slip. The distribution of slip is actually non–
uniform along a fault, however detailed source models are not necessary when
calculating the far–field stress distributions (KING and COCCO, 2001), thus uniform
average slip distributions are adequate for all the Coulomb calculations. Fault
lengths were determined from geology, morphology, detailed macroseismic descrip-
tions, and the spatial distribution of aftershocks (PAPAZACHOS et al., 2001). These
fault lengths were compared with the values derived by the following scaling law
suggested for the broader area of Greece (PAPAZACHOS, 1989):
log L¼0:51 M1:85;ð2Þ
that gives the fault length, L, as a function of the main-shock magnitude, M
w
, and
found to be in a good agreement. The average displacement of each event was
computed from the events static moment assuming a rigidity of 33 GPa, from the
fault lengths, and from the depth of the seismogenic layer (that give the fault areas).
Figures 7a–i are snapshots of DCFF at a depth of 8 km. This depth was chosen to
be several kilometers above the locking depth (15 km) in the evolutionary model.
This is in agreement with KING et al. (1994b) who found that seismic slip peaks at
mid–depths in the seismogenic zone. In these figures dark regions denote negative
changes and inferred decreased likelihood of fault rupture, and are called stress
shadows (HARRIS and SIMPSON, 1993, 1996), while light regions represent positive
DCFF and the increased likelihood of fault rupture, and are called stress bright zones.
Shadow zones and bright zones are specific to strike, dip, and rake of the fault that
experiences the DCFF . We will show that most of the larger earthquakes occurred in
bright zones. Moreover, moderate–size shocks with faulting similar to the type for
Figure 7
Stress evolution in the area of the southern Aegean since 1904. Coulomb stress is calculated for normal
faults at a depth of 8.0 km. The stress pattern is calculated for the faulting type of the next large event in
the sample. Changes are denoted by the gray scale at bottom (in bars). Contour lines separate positive from
negative DCFF . Fault–plane solutions are plotted as lower–hemisphere equal–area projections. The
occurrence year of each event is given on top of the focal spheres. The major normal faults of the area
associated with the occurrence of known strong events are designated white. The fault for which the stress
field is calculated in each stage of the evolutionary model is designated black. Black stars represent
moderate magnitude events (two sizes for two magnitude ranges, 5.0 M5.4 and 5.5 M5.9,
respectively) while open stars represent the events of 6.0 M6.4. (a) Coseismic Coulomb stress changes
associated with the 1904 Samos event. (b) Stress evolution until just before the 1922 Zakros event. (c)
DCFF just before the 1928 Torbali and (d) 1933 Kos events. (e) State of stress before the occurrence of 1948
Karpathos event and (f) before the 1955 W. Buyuk earthquake. (g) Stress evolution just until and (h) after
the 1956 Amorgos main shock. (i) DCFF just before the the 1956 Dionisades earthquake.
c
794 E. Papadimitriou et al. Pure appl. geophys.,
Vol. 162, 2005 Seismicity Variations in the Southern Aegean 795
which the stress calculations are performed, were also located in stress–enhanced
zones.
The coseismic displacements in the seven larger earthquakes in the southeastern
Aegean area since the beginning of the 20
th
century, i.e., those with M 6.5
(Table 2), are included in the evolutionary stress model. The faults are simplified and
approximated by rectangular shapes. All of these events involve normal faulting with
fault planes oriented either in an E–W or N–S direction. At each stage, DCFF is
calculated for a specific fault plane solution, that of the next inspected event. The
changes in stress are presented for the whole study area.
1904, Samos earthquake (M6.8). Figure 7a shows the coseismic stress changes
associated with this event, with a rupture length of 40 km and an estimated
displacement of 88 cm. Rupture is taken to extend throughout the seismogenic zone,
i.e., from 3 to 15 km. The stress field was computed according to its fault plane
solution (strike ¼91, dip ¼45, rake ¼)115). The earthquake created a shadow
zone in the northern, central, and southern parts of our study area and bright zones
to the east and west. We expect these stress changes to affect the occurrence of future
events.
1922, East Crete earthquake (M6.8). This event occurred east of Crete Island on
a N–S trending normal fault, inside an area of positive static stress changes (Fig. 7b)
calculated according to its fault plane solution (strike ¼14, dip ¼47,
rake ¼)98). With a rupture length of 40 km an average slip of 94 cm was
estimated. The epicenters of smaller magnitude events that occurred until its
occurrence are also located in bright zones.
1928, Torbali earthquake (M6.5). Figure 5c shows the accumulated stress
changes calculated according to the fault plane solution (strike ¼83, dip ¼45,
rake ¼)115) and just before the occurrence of the 1928 Torbali earthquake. Its
rupture zone as well as the events that occurred during 1923–1928 is located inside a
region of positive DCFF .
1933, Kos earthquake (M6.6). In Figure 5d the accumulated Coulomb stress
changes just before the occurrence of the 1933 event are shown for faulting in
agreement with its focal mechanism solution (strike ¼65, dip ¼50, rake ¼)90).
In the same figure the events of M 5.0 that occurred from 1929 to 1933 are shown.
Two of them are located in bright zones and the remaining two near the borders
between bright and shadow zones.
1948, Karpathos earthquake (M6.9). This earthquake is associated with a N–S
trending normal fault and is located inside an area of positive static stress changes
(Fig. 7e) calculated according to its fault plane solution (strike ¼185, dip ¼47,
rake ¼)98). Intense activity during 1934–1948, with three events of M 6.0, is
well correlated with the areas of positive DCFF .
1955, Boughiouk Menteres earthquake (M6.9). Figure 7f shows the accumulated
Coulomb stress changes just before the 1955 earthquake calculated according to its
796 E. Papadimitriou et al. Pure appl. geophys.,
Table 2
Rupture models for earthquakes with M 6.5 that occurred in the study area during the 20th century and are included in the evolutionary model
Date Time Latitude
(uN)
Longitude
(kE)
Depth (km) L (km) u (cm) M Mechanism Ref.
Strike Dip Rake
1904, Aug. 11 06:08:30 37.71 26.85 3–15 40 88 6.8 91 45 )115 1, 2
1922, Aug. 13 00:09:54 35.16 26.49 3–15 40 94 6.8 14 47 )98 1, 2
1928, Mar. 31 00:29:47 38.18 27.45 3–15 30 43 6.5 83 45 )115 1, 2
1933, Apr. 23 05:57:37 36.75 27.19 3–15 30 67 6.6 65 50 )90 1, 2
1948, Feb. 9 12:58:13 35.76 27.05 3–15 48 118 6.9 185 47 )98 1, 2
1955, July 16 07:07:10 37.715 27.40 3–15 45 110 6.9 91 45 )115 1, 2
1956, July 9 03:11:40 36.82 26.12 3–15 75 500 7.7 65 40 )90 1, 4, 5
1956, July 9 03:24:03 36.58 25.49 3–15 40 112 6.9 50 40 )90 1, 3
1. PAPAZACHOS et al. (2000); 2. PAPAZACHOS et al. (1998); 3. PAPAZACHOS et al. (2001); 4. SHIROKOVA (1972); 5. PACHECO and SYKES (1992).
Vol. 162, 2005 Seismicity Variations in the Southern Aegean 797
fault plane solution (strike ¼91, dip ¼45, rake ¼)115). The epicenters of smaller
magnitude shocks are located inside or at the borders of stress enhanced areas.
1956, Amorgos earthquake (M7.7). Figure 7g shows the state of stress before
this earthquake in respect to the 1904 baseline, calculated according to its fault
plane solution (strike ¼65, dip ¼40, rake ¼)90). The main shock as well as
the stronger events (M 6.0) during 1956 is also shown in this figure. Modeled
stress evolution including the coseismic stress changes associated with the M7.7
main shock is shown in Figure 7h. The Santorini earthquake (1956b) that
occurred a few minutes afterward is located in an area of increased positive stress
changes due to the main-shock occurrence. In Figure 7i the coseismic displace-
ment of the Santorini event is also included and the stress field is calculated
according to the fault plane solution of Dionisades (1956c) earthquake
(strike ¼10, dip ¼47, rake ¼)98) which is located inside the area of positive
static stress changes. The correlation of the location of the last two earthquakes
with positive DCFF evidences their possible consequent triggering.
State of stress in 1983. During 1956–1983 no large event (M 6.5) occurred in
our study area. The snapshot in Figure 8a indicates the evolved state of stress as
of 1983, which differs from that of Figure 7g in that it includes the stress
accumulation caused by the 27 additional years of tectonic loading. Hence,
stresses are derived for a typical E–W trending normal fault for the area
(strike ¼65, dip ¼40, rake ¼)90). The focal mechanisms of the events
occurring between 1957 and 1983 of M 5.5 with this faulting type are also
shown (Table 3). Their epicenters are located inside bright zones or near the
borders of bright and shadow zones. The stress field in Figure 8b differs from that
in Figure 8a in that it is computed for N–S trending normal faults (strike ¼185,
dip ¼47, rake ¼)98). Two events of such type of faulting that occurred 1983
are situated in an area of negative DCFF .
State of stress in 2003. The Coulomb stress evolutionary calculations
continued until 2003 with an additional 20 years of tectonic loading on the
major faults. Figure 8c depicts the evolved stress field calculated for E–W
trending normal faults and likewise Figure 8c for N–S trending ones. All but one
of the events depicted in Figure 8c are inside bright zones while in Figure 8d the
model failed to predict one of the two events of this period.
5. Discussion and Conclusions
The variation in the spatio–temporal occurrence of moderate and large
earthquakes and its correlation with changes in the stress field comprises important
evidence in the study of earthquake generation. From this point of view we examined
the occurrence mode of moderate earthquakes both before and after the 1956
Amorgos main shock, which is the larger event (M7.7) that occurred in the Aegean
798 E. Papadimitriou et al. Pure appl. geophys.,
area during the instrumental era. The seismicity inside circular areas centered on the
main shock epicenter, with radii equal to 100, 150 and 200 km taken into account,
and patterns both in time and space were pursued. The choice of the radii was rather
Figure 8
Same as Figure 7 for (a) state of stress as of 1983 computed for the faulting type of the E–W trending
normal faults and (b) for N–S trending faults. The evolved stress field is calculated until 2003 for E–W
trending normal faults (c) and for N–S trending faults (d). Seismicity is also depicted. Fault plane solutions
of events with M 5.5 are shown as lower–hemisphere equal–area projections.
Vol. 162, 2005 Seismicity Variations in the Southern Aegean 799
arbitrary, and perhaps the observed patterns would be more striking if an additional
effort was made to define the most appropriate radii, although the accuracy in
epicentral determination does not allow a very detailed investigation.
From the temporal variation of the occurrence frequency it was found that the
seismicity behaved in a converse way near the main shock epicenter, that is at
distances up to 100 km, rather than in the longer distances, and changed
dramatically after the occurrence of the 1933 main shock that is associated with
the nearby Kos normal fault. More specifically, from 1933 until 1956 the circular
area with a radius equal to 100 km was completely quiescent. On the contrary,
seismic excitation was observed at longer distances during 1933–1956. This excitation
is more obvious when the occurrence rate or the spatial distribution of the events
inside the annuli with radii 100–150 km and 150–200 km was examined. It is worthy
to note that lack of activity for events of M 6.0 was observed during 1920–1933 at
distances 100–200 km, when several events of M 6.0 occurred during 1933–1956,
due to the successive possible triggering of the stronger events as is evidenced by the
stress evolutionary model. The changes in the occurrence rate are more striking for
distances of 100–150 km, where the activated faults are located, and magnitude range
5.5 M5.9. These values correspond to an almost twice larger rupture extent and
2 orders smaller magnitude than the main shock. This intermediate magnitude
seismicity increase is also associated with areas of positive DCFF created during this
period. This finding is in accordance with DUand SYKES (2001) who found the most
prominent changes of the rates of moderate–size events before and after the Landers
Table 3
Information on the fault plane solutions of earthquakes with M 5.5 that occurred in the study area during
1957–2003 (data from PAPAZACHOS et al., 1998; Harvard solution)
Date Time Latitude
(uE)
Longitude
(kN)
M Mechanism
Strike Dip Rake
1961, Feb. 23 21:45:54 36.7 27.1 5.6 263 53 )111
1962, Apr. 28 11:18:59 36.1 26.8 5.8 86 50 )90
1962, Apr. 28 12:43:49 36.1 26.9 5.6 86 50 )90
1968, Oct. 31 03:22:14 36.6 27.0 5.7 86 50 )90
1968, Dec. 5 07:52:11 36.6 26.9 6.0 73 52 )100
1975, Sep. 22 00:45:01 35.38 26.35 5.5 18 50 )97
1979, July 23 11:41:54 35.43 26.33 5.6 186 68 )119
1989, Apr. 27 23:06:52 37.06 28.03 5.5 73 52 )100
1989, Apr. 28 13:30:20 37.06 28.01 5.5 73 52 )100
1990, Aug. 28 20:21:21 36.38 27.13 5.5 86 50 )90
1992, Apr. 30 01:44:40 35.10 26.60 5.7 172 38 )106
1992, Nov. 6 19:18:10 38.19 27.05 6.0 238 85 )167
1993, Aug. 26 10:03:56 36.75 28.06 5.6 73 52 )100
1996, July 20 00:00:39 36.07 27.459 6.2 196 38 )102
800 E. Papadimitriou et al. Pure appl. geophys.,
earthquake, within a circular area centered on the epicenter of the main shock with a
radius of about 160 km. This anomalous activity before the Landers event was
largely confined to a stress–enhanced zone with positive changes in CFF well above
0.01 MPa since 1812.
It is widely accepted during the last decade that accelerating moment release
takes place before large events in areas considerably larger than the main shock
fault. In the present study, it was ascertained that the frequency of moderate
events does increase during several years before the main shock and is
concentrated around the ends of the forthcoming rupture, defining thus simulta-
neously a quiescent period for the focal region. The doughnut pattern was created
by the activation of the surrounding faults that reached their critical stress and
became active before the activation of the Amorgos fault. KATO et al. (1997)
demonstrated that the regional stress relaxation due to the preseismic sliding tends
to lower the seismic activity, leading to the appearance of precursory seismic
quiescence. Their model explains previous observations (OHTAKE, 1980; MOGI,
1985) stating that the precursory seismic quiescence often appears in a significantly
wider area than the source area of the main shock; that the duration of quiescence
is larger for a larger main shock, and that the duration of quiescence is a few
years to a few decades for great earthquakes of M 8. This model development
leads to the indication that the stress variation due to the aseismic sliding from the
upper aseismic zone as well as from the lower aseismic zone into the central part
of the seismogenic zone may explain these seismicity patterns both in the
overriding continental plate and in the subducting oceanic plate (KATO and
HIRASAWA, 1999). The rise of seismic activity is due to the involved physical
process and concerns the dimensions of the stress field that is in a critical stage at
distances much greater than the fracture dimensions. Contemporarily, the
observed quiescence corresponds spatially with the fault extent of the ensuing
main shock. The model of KING and BOWMAN (2003) that links Coulomb stress
changes and the regional seismicity explains well why pre-event seismicity should
appear at a distance from the future epicenter and the relative quiescence
following major earthquakes. In this model which concurs with the findings of the
present study, the epicentral region is quiescent over much of the earthquake cycle,
as is commonly observed for real earthquake sequences, and when the activity
recommences prior to a future event, it occurs around the quiescent region to form
a ‘‘Mogi Doughnut’’ (MOGI, 1969, 1981). In our case increased seismic activity
commenced in 1933, comprising strong and intermediate magnitude events that
occurred in some distance from the Amorgos fault, and which are well correlated
with stress-enhanced areas in each stage of the evolutionary model. During the last
52 years (1957–2004) no earthquake with M 6.5 occurred in our study area in
contrast with the previous interval of equal duration (1904–1956) which comprises
six events of M 6.5. The current state of stress explains this remarkable
quiescence in larger magnitude earthquakes given that the major faults in the
Vol. 162, 2005 Seismicity Variations in the Southern Aegean 801
study area, which are associated with the occurrence of such events, are currently
in stress shadows.
Acknowledgements
Two of the authors (E. P. and V. K.) are indebted to Lynn Sykes whose impetus
motivated them to work on the subject. The stress tensors were calculated using the
DIS3D code of S. Dunbar, which was later improved by ERIKSON (1986) and the
expressions of G. Converse. The GMT system (WESSEL and SMITH, 1995) was used to
plot the figures. The comments of two anonymous reviewers and the editorial
assistance of Kunihiko Shimazaki are greatly appreciated. This work was partially
supported by the project EPAN–M.4.3.6.1 funded by the General Secretariat of
Research and Technology of Greece. Geophysics Department contribution 641.
R
EFERENCES
BOWMAN, D. D., OUILLON, G., SAMMIS, C. G., SORNETTE, A., and SORNETTE, D. (1998), An Observational
Test of the Critical Earthquake Concept. J. Geophys. Res. 103, 24,359–24,372.
DENG, J. and SYKES, L. R. (1997), Evolution of the Stress Field in Southern California and Triggering of
Moderate–size Earthquakes: A 200–year Perspective. J. Geophys. Res. 102, 9859–9886.
DU, W. –X. and SYKES, L. R. (2001), Changes in Frequency of Moderate–size Earthquakes and Coulomb
Failure Stress before and after the Landers, California, Earthquake of 1992. Bull. Seismol. Soc. Am. 91,
725–738.
ELLSWORTH, W. L., LINDH, A. G., PRESCOTT, W. H., and HERD, D. G. (1981), The 1906 San Francisco
earthquake and the seismic cycle.InEarthquake Prediction: An International Review (eds. Simpson, D.
W. and Richards, P. G.), (AGU, Washington, D. C. 1981), pp. 126–140.
ERIKSON, L. (1986), User’s Manual for DIS3D: A Three–dimensional Dislocation Program with Applications
to Faulting in the Earth, Masters Thesis, Stanford Univ., Stanford, Calif., 167 pp.
FEDOTOV, S. A. (1965), Regularities in the Distribution of Strong Earthquakes in Kamchatka, the Kuriles,
and Northeastern Japan, Akad. Nauk USSR Inst. Fiz. Zeml.: Trudy 36, 223–246.
HARRIS, R. A. and SIMPSON, R. W. (1993), In the Shadow of 1857: An Evaluation of the Static Stress
Changes Generated by the M8 Ft. Tejon, California, Earthquake, EOS Trans. AGU, 74(43), Fall Meet.
Suppl. 427.
HARRIS, R. A. and SIMPSON, R. W. (1996), In the Shadow of 1857: The Effect of the great Ft. Tejon
Earthquake on Subsequent Earthquakes in Southern California, Geophys. Res. Lett. 23, 229–232.
KATO, N. and HIRASAWA, T. (1999), The Variation of Stresses due to Aseismic Sliding and its Effect on
Seismic Activity, Pure Appl. Geophys. 155, 425–442.
KATO, N., OHTAKE, M., and HIRASAWA, T. (1997), Possible Mechanism of Precursory Seismic Quiescence:
Regional Stress Relaxation due to Preseismic Sliding, Pure Appl. Geophys. 150, 249–267.
KING, G. C. P. and COCCO, M. (2001), Fault Interaction by Elastic Stress Changes: New Clues from
Earthquake Sequences, Adv. Geophys. 44, 1–38.
KING, G. C. P. and BOWMAN, D. D. (2003), The Evolution of Regional Seismicity between Large
Earthquakes, J. Geophys. Res. 108, B2, 2096, doi:10.1029/2001JB000783.
KING, G. C. P., STEIN, R. S., and LIN, J. (1994a), Static Stress Changes and the Triggering of Earthquakes,
Bull. Seismol. Soc. Am. 84, 935–953.
KING, G., OPPENHEIMER, D., and AMELUNG, F. (1994b), Block Versus Continuum Deformation in the
Western United States, Earth Planet. Sci. Lett. 128, 55–64.
802 E. Papadimitriou et al. Pure appl. geophys.,
KNOPOFF, L., LEVSHINA, T., KEILIS–BOROK, V. I., and MATTONI, C. (1996), Increased Long–range
Intermediate–magnitude Earthquake Activity prior to Strong Earthquakes in California, J. Geophys. Res.
101, 5779–5796.
MCCLUSKY, S., BALASSANIAN, S., BARKA, A., DEMIR, C., GEORGIEV, I., HAMBURGER, M., HURST, K.,
KAHLE, H., KASTENS, K., KEKELIDZE, G., KING, R., KOTZEV, V., LENK, O., MAHMOUD, S., MISHIN, A.,
NADARIYA, M., OUZOUNIS, A., PARADISIS, D., PETER, Y., PRILEPI, M., REILINGER, R., SANLI, I.,
SEEGER, H., TEALEB, A., TOKSOZ, M.N., and VEIS, G. (2000), GPS Constraints on Crustal Movements
and Deformations in the Eastern Mediterranean (1988–1997): Implications for Plate Dynamics,
J. Geophys. Res. 105, 5695–5719.
MCKENZIE, D. P. (1978), Active tectonics of the Alpine–Himalayan Belt: The Aegean Sea and Surrounding
Regions, Geophys. J. R. astron. Soc. 55, 217–254.
MOGI, K. (1969), Some Features of Recent Seismic Activity in and near Japan (2). Activity before and after
Great Earthquakes, Bull. Earthquake Res. Inst., Univ. Tokyo 47, 395–417.
MOGI, K. (1981) Seismicity in western Japan and long–term earthquake forecasting,InEarthquake
Prediction: An International Review (eds. Simpson, D.W. and Richards, P.G.), (AGU, Washington, D.C.
1981), pp. 43–51.
MOGI, K., Earthquake Prediction (Academic Press, Tokyo 1985).
NALBANT, S. S., HUBERT, A., and KING, G. C. P. (1998), Stress Coupling between Earthquakes in
Northwestern Turkey and the North Aegean Sea, J. Geophys. Res. 103, 24469–24486.
OKADA, Y. (1992), Internal Deformation due to Shear and Tensile Faults in a Half–space, Bull. Seismol. Soc.
Am. 82, 1018–1040.
OHTAKE, M. (1980), Earthquake Prediction Based on the Seismic Gap with Special Reference to the 1978
Oaxaca, Mexico Earthquake, Rept. Nat. Cent. Disaster Prev. 23, 65–110 (in Japanese).
PACHECO, J. F. and SYKES, L. R. (1992), Seismic Moment Catalog of Large Shallow Earthquakes, 1900 to
1989, Bull. Seismol. Soc. Am. 82, 1306–1349.
PAPAZACHOS, B. C. (1989), Measures of earthquake size in Greece and surrounding areas, Proc. of the 1
st
Scient. Conf. of Geophysics, Geophys. Soc. of Greece, April 19–21, 1989, Athens, 438–447.
PAPAZACHOS, B. C. and COMNINAKIS, P. E. (1970), Geophysical Features of the Greek Island Arc and
Eastern Mediterranean Ridge, Com. Ren. Des Sceances de la Conference Reunie a Madrid, 1969 16, 74–
75.
PAPAZACHOS, B. and PAPAZACHOS, C. (2000), Accelerated Preshock Deformation of Broad Regions in the
Aegean Area, Pure Appl. Geophys. 157, 1663–1681.
PAPAZACHOS, B. C. and PAPAZACHOU, C. (2002), The earthquakes of Greece, Ziti Publ., Thessaloniki,
Greece, 317 pp.
PAPAZACHOS, B. C., KIRATZI, A. A., and KARAKOSTAS, V. G. (1997), Toward a Homogeneous Moment
Magnitude Determination in Greece and Surrounding Area, Bull. Seismol. Soc. Am. 87, 474–483.
PAPAZACHOS, B. C., PAPADIMITRIOU, E. E., KIRATZI, A. A., PAPAZACHOS, C. B., and LOUVARI,E.K.
(1998), Fault Plane Solutions in the Aegean Sea and the Surrounding Area and their Tectonic
Implications,Boll. Geof. Teor. Appl. 39, 199–218.
PAPAZACHOS, B. C., COMNINAKIS, P. E., KARAKAISIS, G. F., KARAKOSTAS, B. G., PAPAIOANNOU, Ch. A.,
PAPAZACHOS, C. B., and SCORDILIS, E. M. (2000), A Catalogue of Earthquakes in Greece and Surrounding
Area for the Period 550 BC–1999.http://geohazards.cr.usgs.gov /iaspei/europe/greece/the/catalog.htm.
PAPAZACHOS, B. C., MOUNTRAKIS, D. M., PAPAZACHOS, C. B., TRANOS, M. D., KARAKAISIS, G. F., and
SAVVAIDIS, A. S. (2001), The Faults that Caused the Known Strong Earthquakes in Greece and
Surrounding Areas during 5
th
century B. C. up to Present,2
nd
Conf. Earthq. Eng. and Eng. Seismol., 28–
30 September 2001, Thessaloniki 1, 17–26.
SCHOLZ, C., The Mechanics of Earthquakes and Faulting (Cambridge University Press, Cambridge 1990)
439 pp.
SCORDILIS, E. M., KARAKAISIS, G. F., KARAKOSTAS, B. G., PANAGIOTOPOULOS, D. G., COMNINAKIS, P. E.,
and PAPAZACHOS, B. C. (1985), Evidence for Transform Faulting in the Ionian Sea: The Cephalonia Island
Earthquake Sequence, Pure Appl. Geophys. 123, 388–397.
SHIROKOVA, E. (1972), Stress pattern and probable motion in the earthquake foci of the Asia-Mediterranean
seismic belt, in elastic strain field of the earth and mechanisms of earthquake sources. In (L. M. Balakina
et al., eds.), Nauka, Moscow, 8.
Vol. 162, 2005 Seismicity Variations in the Southern Aegean 803
STEKETEE, J. A. (1958), On Voltera’s Dislocations in a Semi–infinite Elastic Medium, Can. J. Phys. 36, 192–
205.
SYKES, L. R. and JAUME
´, S. C. (1990), Seismic Activity on Neighboring Faults as a Long–term Precursor to
Large Earthquakes in the San Francisco Bay Area, Nature 348, 595–599.
WESSEL, P. and SMITH, W. H. F. (1995), New Version of the Generic Mapping Tools Released, EOS, Trans.
Am. Geophys. U. 76, 329.
(Received July 25, 2003, accepted June 29, 2004)
To access this journal online:
http://www.birkhauser.ch
804 E. Papadimitriou et al. Pure appl. geophys.,
