Geoph.vs. J. Int. (1997) 131,478-454
SPECIAL SECTION-ASSESSMENT
OF S C H E M E S F O R E A R T H Q U A K E P R E D I C T I O N
Costs and benefits of earthquake prediction studies in Greece
Stathis C. Stiros
Department of Civil Engineering, Patras Universrtj, Patra 26500, Greece E-mail strros(a hol gr
Accepted 1997 August 28. Received 1997 August 22; in original form 1997 January 31
SUMMARY
For the following reasons, a (hypothetical) successful earthquake prediction in Greece
would be of, at best, limited benefit to society. (1) On average, less than 5 per cent of
sizeable earthquakes (that is magnitude greater than 4.5) cause significant damage or
loss of life. (2) Organized evacuation of urban centres is unlikely to be successfully
accomplished, because the public lacks confidence in the authorities and cannot be
expected to respond promptly; panic and other undesirable side-effects can also
be anticipated. ( 3 ) The lead time between a short-term prediction and earthquake
occurrence is too short for most actions aimed at reducing or eliminating primary or
secondary earthquake effects; in any case, most such actions would be superfluous if
appropriate longer-term preparedness plans were implemented. (4) Prediction of
‘an earthquake’ is not an appropriate objective in an area such as Greece, which
experiences complicated and long seismic sequences consisting of several destructive
events, and large earthquakes with anomalous meizoseismal areas. ( 5 ) The seismic
death toll in Greece is relatively small (less than 10 people per year over the last
40 years), and due to recent changes in building styles and construction practices,
current morbidity is mainly associated with the failure of multistorey buildings. This
death toll could be more effectively eliminated by identifying weaknesses and structural
intervention than by earthquake prediction. Hence, earthquake prediction in Greece,
even if it were scientifically feasible, would not be cost-effective; alternative use of
funding could be expected to save more lives with much greater certainty. Over the past
15 years, the VAN group’s research on earthquake prediction has absorbed a substantial
fraction of the resources devoted to earthquake research and protection in Greece.
However, the VAN method has not advanced the nation’s policy on earthquake protection planning, its results continue to be widely questioned by both the Greek and
international scientific communities, and the underlying model is not commensurate
with currently accepted thinking on earthquake generation and tectonophysics. Thus,
VAN’S methods can be regarded as basic research rather than as an operational method
for the reduction of seismic risk.
Key words: earthquake prediction, Greece.
1
INTRODUCTION
Earthquakes with magnitudes higher than 6.5 or even 7.0
are not unusual in Greece (Makropoulos & Burton 1981;
Ambraseys & Jackson 1990), but for three reasons cause
relatively little damage. (1) The majority of earthquakes
occur beneath the sea; seismic waves are therefore attenuated
by the time they reach inhabited areas. (2) The geological
structure of the areas causes rapid attenuation of seismic
waves (Papazachos & Papazachou 1989; Ambraseys & Jackson
1990); in western Greece, in particular, alternating sheets of
478
carbonates and evaporites dramatically limit the propagation
of seismic waves in a selective fashion (Stiros 1994; Stiros
et al. 1994). ( 3 ) Special techniques for the construction of
earthquake-resistant houses have traditionally been widely
used in Greece and adjacent regions (Stiros 1995a). Since 1928,
when the town of Corinth was totally destroyed by an earthquake, building codes for earthquake-resistant houses have
been gradually introduced nationally.
Between 1978 and 1986, several moderate earthquakes hit
four main urban centres in Greece, including Athens and
Thessaloniki, the two major cities, causing much destruction
1997RAS
Earthquake prediction studies in Greece
and many casualties (1978 Thessaloniki, M , =6.5, 45 people
killed due to the collapse of a multistorey building; 1980 Volos,
M,=6.5, no deaths; 1981 Corinth and Athens, M,=6.7, 20
deaths; 1986 Kalamata, M, = 6.0, 20 deaths; Papazachos &
Papazachou 1989). As a result, scientific and social interest in
earthquakes increased, and methods for reducing seismic risk
were introduced. EPPO, the Greek Organization for AntiSeismic Protection (OASP in Greek) was established, and
new anti-seismic building codes were introduced. This period
was also notable for studies that combined analyses based on
synthetic seismograms with detailed field mapping of surface
deformation (e.g. Jackson et al. 1982).
The 1980s were also marked by the advent of earthquake prediction research in Greece. Wyss & Baer (1981)
hypothesized the existence of a seismic gap in the SW part of
the Hellenic arc, and claimed that there was a high probability
of a strong earthquake in this area in 1980-1990. However, this
earthquake did not occur. Their theory was adopted by
Papadopoulos (1988a,b), who announced that an earthquake
was to be expected in this area in 1986. A public prediction of
such an earthquake was made in the Greek daily newspaper
Elephtherotypia (25 August 1986) a few weeks before the
Kalamata earthquake of 1986 September 13; Papadopoulos
(1988a,b) claimed that this was the expected earthquake.
Papadopoulos also claimed a successful prediction of the
Aigion (1995 June 15) earthquake, based on the hypothesis that
strong earthquakes in Central Macedonia (such as that of 1995
May 6) are followed by earthquakes in the Gulf of Patras, near
Aigion (Papadopoulos 1996).
Constantinides (1987) reported an anomalous (up to a few
metres) change in groundwater levels 5-30 days before, and at
distances of 30 km or more from the seismogenic fault of the
1980 Almyros earthquake in Central Greece. Asteriadis &
Livieratos (1989) and Asteriadis & Zioutas (1990) proposed the
existence of a correlation between centimetre-wide fluctuations
in groundwater level and small earthquakes in the epicentral
area of the 1978 Thessaloniki earthquake.
Other scientists in Greece have made earthquake predictions
on the basis of changes in telluric currents (Thanassoulas 1991;
Ifantis et al. 1993) and electric currents (Thanassoulas &
Tselentis 1993). However, the most notable earthquake prediction efforts have been those of P. Varotsos and co-workers,
known as ‘VAN. These workers argue that they have recorded
ground electric signals preceding and genetically related to
earthquakes (‘seismic electric signals’ or SES; Varotsos &
Alexopoulos 1984; Varotsos & Lazaridou 1991). It was and
is claimed that such signals can be analysed to give warning
of the magnitudes of earthquakes that will occur in specified
areas within a few hours (as was claimed in the early 1980s)
to a few weeks (as has been claimed in recent years; see
Stavrakakis & Drakopoulos 1996) after the occurrence of
SES (Varotsos, Alexopoulos & Nomikos 1981; Varotsos,
Alexopoulos & Lazaridou 1993). From its inception, this
group’s work has been presented and/or interpreted as a readyto-use, operational and successful method for short-term
earthquake prediction. Over the period from 1981, the VAN
group has issued several tens of announcements of SES,
accompanied by warnings of impending earthquakes, and
these have attracted both national and international attention.
This high public profile has been reflected in high levels of
funding. The interested reader is referred to Anagnostopoulos
(1995) for a fuller account of these matters than can be pre0 1997 RAS, GJI 131,478434
479
Table 1. Distribution of funding for seismological a n d related studies
in Greece, 1981-1989, from the Earthquake Protection and Planning
Organisation (EPPO/OASP).
Open procedures,
proposals, peer review,
contracts, etc?
VAN
Seismology/ Geology
(incl. seismological networks)
Public awareness, education
Studies for antiseismic
construction, building codes, etc.
40.2%
No
Yes
3.2%
10.4%
(internal activity)
Yes
46.2%
Source: Anagnostopoulos (1995)
sented here. Table 1 compares the distribution of EPPO/OASP
funding to VAN during the period 1981-89 with that available
for related objectives. However, Anagnostopoulos (1995) also
states that following a change in government and in procedures
at EPPOIOASP, the VAN group did not submit a proposal to
the peer-reviewed system which was in place during the period
following.
Although their work has been well publicised in the
popular media, and a number of invitation-only meetings
have been held to discuss their method, the VAN group have
presented their work at relatively few open scientific meetings
within Greece itself [for example this author is aware of
contributions only at the Conference of the Geological
Society of Greece on Geology and Earthquakes, May 1984
(Varotsos 1987), and at the IASPEI congress of August 1997,
held in Thessaloniki]. Thus scientific debate on VAN in
Greece has been minimal (Papazachos et al. 1987;
Drakopoulos & Latousakis 1987; Stavrakakis, Drakopoulos
& Latousakis 1990; the contributions of Drakopoulos and
colleagues cited by Stavrakakis & Drakopoulos 1996).
Early discussions on VAN by international scientists (e.g.
Burton 1985) had almost no impact in Greece. Scientific
criticism of VAN ranged between the endpoints of total
rejection (Papazachos et al. 1987; Drakopoulos &
Latousakis 1987) and the hope that VAN might in the
future develop a useful method of prediction, on the basis
that the signals they were observing apparently had some
relation to earthquakes (Anagnostopoulos 1995).
With one exception, VAN’S claims of successful predictions
have not been officially recognized as accurate or reliable
by Greek authorities. Following the 1995 May 6 Central
Macedonia earthquake, a statement acknowledging a successful prediction was made by the Head of EPPO, who until
then had been associated with VAN (Papanikolaou 1993;
Masood 1995; Anagnostopoulos 1995). This statement,
however, had little impact, because this earthquake caused
only insignificant damage, and the announcement was not
followed by any preparedness measures. On the other hand,
the Greek government has supported VAN’s research by
establishing a Centre for Prediction and Prognosis of
Earthquakes in Athens under an Open Agreement with the
Council of Europe, and establishing at the University of
Athens in 1994 an independent and autonomously administrated University Research Institute for the Physics of the
Earth’s Crust devoted to VAN’s activities. Fixed funding
from government sources has been allocated to this
Institution (Decrees 322/1994 and 51/1996 of the President
of the Hellenic Republic).
480
2
S. C. Stiros
E V A L U A T I O N OF V A N
A critical evaluation of a theory or method for earthquake
prediction should focus on whether that theory or method is
(1) scientifically sound, (2) useful, and (3) cost-effective. In the
present context, ‘scientifically sound’ means (la) that data
are reliable and available for evaluation, (lb) that the results
are consistent, statistically significant and supported by
independent evidence, and, ideally, (lc) that there exists a
physical explanation, both qualitative and quantitative. In
the absence of a full physical explanation (lc), it should at
least be demonstrated that the results do not violate known
physical laws. After evaluation, it should be possible to
assign any method for earthquake prediction to one of the
following categories: (A) scientifically sound and successful, as
well as cost-effective and useful; (B) scientifically sound and
promising, but for various reasons (cost, etc.) not immediately
applicable; (C) an interesting but experimental method of
unproven value; (D) a method of limited scientific interest
which is unlikely to lead to applicable results; (E) a method of
no scientific value.
Discussion of VAN has focused on points (la) [Stavrakakis
& Drakopoulos 1996; Geller 1996; Mulargia & Gasparini 1992;
special issue of Geophys. Res. Lett., 1996,23(1 I), etc.] and (lb)
(Papazachos et al. 1987; Geller 1996; Drakopoulos’ various
publications, etc.). Point (2) was addressed by Geller (1996) for
the case of the Kobe earthquake, and partly by Yoshii (1993)
for the case of a 1988 earthquake in SW Greece. Point (3) has
almost been ignored, and point (lc) has also not been examined
much (Burton 1985; Papazachos et al. 1987; Bernard 1992;
Gruszow et al. 1996; Bernard et al. 1997).
A major objective of this paper is hence to present a number
of comments on the scientific context of this method.
(1) Earthquakes occur in various tectonic and tectonophysical environments (depending on the focal depth, etc.)
and are associated with a variety of rupture processes, as the
varying patterns of seismic sequences indicate (earthquake
sequences may consist of one single shock, or several multiple
shocks; they may or may not be accompanied by foreshocks or
aftershocks). This casts doubt on whether all earthquakes will
be preceded by, or by the same type of, electric signals
(Galanopoulos 1981). In fact, the observed differences in
the character of SES received (see Varotsos et al. 1993) may
indicate different origins for the so-called ‘SES’.
(2) Seismicity in Greece is neither spatially nor temporally
uniform, and earthquakes sometimes occur in areas that have
been dormant for centuries, such as the Central Aegean or
Western Macedonia (Galanopoulos 1981; Stiros 1995b, 1997).
Thus, even if the so-called SES are indeed related to the
earthquake generation process, there is no reason to believe
that the empirically compiled ‘SES selectivity maps’ (maps
indicating the assumed areas of origin of SES as recorded at
certain VAN stations, see Varotsos et al. 1993), based on
observations covering just a few years, can be reliable.
(3) Some authors have argued that high-conductivity
paths such as deep-rooted dykes and faults can connect the
sources and receivers of assumed pre-seismic electric signals
(e.g. Utada 1993; Labeyrie 1993). However, only for the case of
the Keratea station (KER) near Athens has the existence of a
deep-seated dyke been discussed; there is no evidence of faults
connecting major seismic sources and VAN stations (Fig. 1;
Papanikolaou 1993). No correlation between the ‘tectonic
I
I
I
Figure 1. (a)-(c) Selectivity maps (shaded areas) of the Ioannina
(IOA), Assiros (ASS) and Keratea (KER) stations of VAN (marked by
solid dots) and major faults (broken lines) according to Varotsos rt al.
(1993) and Papanikolaou (1993). (d) Dotted area east of the surface
trace of the Pindus Thrust (solid line) corresponds to a Mesozoic
evaporitic basin, characterized by a detachment, a low electrical conductivity layer, up to 1.5 km thick, at a depth of approximately 5 km.
See text for references and details. Shaded areas marked TW (‘tectonic
windows’) indicate deep-rooted rocks cropping out at the surface. The
KER station of VAN near Athens is located in such an environment.
Based on data from Bornovas & Rondogianni (1983). No evidence of
high electrical conductivity paths connecting stations IOA and ASS
with the proposed seismic sources exists on the grounds of lithology
and tectonics. This is especially clear for Western Greece, for electric
signals from the various seismogenic volumes should cross the lowconductivity detachment layer before they reach the surface [compare
with (a)]. On the contrary, KER station could be sensitive to certain
seismic sources on the grounds of tectonophysics and ideas proposed
by the VAN group and supporters. However, in a region where seismicity is neither spatially nor temporally uniform, it is difficult to
imagine that observations of a few years could lead to the compilation
of reliable ‘selectivity maps’ relating seismic sources and recording
stations.
windows’ (TW in Fig. Id; bulges of the crystalline basement
cropping out at the surface) and the selectivity areas has ever
been claimed, even though these would appear to represent
ideal high-conductivity paths for SES. Papanikolaou (1993),
however, noticed a strong correlation between the distribution
of sedimentary facies belts (geotectonic zones) and VAN
selectivity maps (Fig. 1). This result is especially important for
the Ioannina (IOA) station of VAN in NW Greece (see Fig. l),
where most of theVAN signals were recorded. This station and
most of its selectivity map can be correlated with carbonate
overburden above a detachment zone composed of evaporites
(salt, gypsum and breccia, see IGRS & IFP 1966; Underhill
1989; Stiros et al. 1994). Detailed geophysical studies of a
salt dome in the Ioannina area revealed that the electric
conductivity of this salt is low and not different from that of the
adjacent flysch (Thanassoulas 1984). Given that the possibility
0 1997 RAS, GJI 131,478484
Eurthquuke prediction studies in Greece
of low-resistivity paths associated with mineralogical effects
can be excluded (references as above), we may conclude that
electric signals from the crystalline basement can be transmitted to the surface only in exceptional circumstances, that is
along faults in which the circulation of fluids can form highelectrical-conductivity zones crossing the evaporitic layer and
its overburden. Since no such faults have yet been documented
in this area, the case that the electric signals recorded at IOA
reflect earthquake-associated effects is, at least at present,
weak.
(4) Arguments for the validity of VAN are based primarily
on claims of a statistically significant correlation between
electrical observations and subsequent earthquakes [see special
issue of Geophys. Res. Lett., 1996 23(11)]. Such claims are
questionable. First, questions concerning the selection of data
(for example the lag time between electric signals and earthquakes, the threshold magnitude of earthquakes, etc.) have
been raised (Papazachos et al. 1987; Geller 1996). Second, the
statistical significance of the correlation is disputed (Kagan &
Jackson 1996; Varotsos et al. 1996). Furthermore, it has been
argued that the correlation of the electrical signals with prior
earthquakes is better than that with subsequent earthquakes
(Drakopoulos & Latousakis 1987; Mulargia & Gasparini
1992). Thus the electric signals observed by VAN, if related to
earthquakes, may be post-seismic rather than pre-seismic
effects.
In view of the above studies, the VAN method cannot
be classified into either of the above categories A or B. At
best, it might be assigned to category C. Taking into account
the criticisms of Papazachos et al. (1987), Geller (1996) and
Stavrakakis & Drakopoulos (1996), it might warrant classification in category D or even E. However, a definitive
evaluation cannot be made until, as a minimum, the following
steps are taken. First, VAN’S continuous electrical records
(as opposed to telegrams or faxes) are made accessible to
independent researchers (Stavrakakis & Drakopoulos 1996).
Second, extensive investigation is undertaken into whether
electric signals similar to those of VAN, but not related to
earthquakes, can be documented. This might be achieved either
by following the practice of Bernard (1992), Grusow et al.
(1996) and Bernard et al. (1997) or by testing VAN techniques
in non-seismic areas, both free and full of industrial noise.
3 BENEFITS OF A SUCCESSFUL
PREDICTION
Two social benefits are widely expected of a successful shortterm earthquake prediction: the saving of lives (Masood 1995)
and a maximum level of preparedness (Anagnostopoulos
1995). Is this realistic? First, we should consider which earthquakes may be destructive and estimate the possibility of
their occurrence and prediction. Second, we should evaluate
the positive and negative effects of a (hypothetical) successful
prediction.
In some countries, even relatively small earthquakes may
have considerable effect. For example, the 1909 November 6
Provence, southern France, earthquake of magnitude 5
destroyed several villages and killed a dozen people. In Greece,
on the other hand, earthquakes with magnitude smaller than
5.5 to 5.8 cause few, if any, fatalities. For example, although the
Kozani-Grevena area was seriously affected by the 1995
0 1997 U S , GJI 131,478484
481
M , = 6.6 earthquake, an earthquake of magnitude 5.6 in 1984
caused only minor damage (Stiros 1995b, 1997). This suggests
that prediction may be socially useful in Greece only for
earthquakes with magnitude higher than about 5.5.
In a few cases, people alarmed by foreshocks have
spontaneously rushed out of their houses and thereby
saved their lives; this occurred before the 1926 Rhodes
earthquake (Papazachos & Papazachou 1989) and the 1995
Kozani-Grevena earthquake (Stiros 1995b). Could a centrally
organized evacuation be equally successful? The answer is
probably no. The main reason is that the above successful
examples of evacuation come from agricultural areas, where
evacuation is simple, and not from urban centres.
A further problem in Greece is that people lack confidence in
the authorities, and therefore cannot be expected to respond
promptly. Even worse, panic might easily follow a public
announcement concerning a predicted earthquake-either its
magnitude (‘they talk about an earthquake of magnitude 6,
because they try to conceal the truth about a magnitude 8
earthquake’) or its epicentre (‘the earthquake will hit not
the specified area, but many other distant areas’) or its
side-effects (‘the earthquake will trigger an eruption of the
Santorini volcano’). These three quotations are examples of
actual responses of local people, which we recorded following
rumours based on VAN predictions of an earthquake in the
southern Aegean. Another factor is fatigue following a long
period of seismic activity, which causes people to ignore risks
(‘I prefer to stay in my house, and not in a tent’). In the case of
the 1956 Amorgos M=7.6 seismic sequence, for instance,
several months of post-seismic activity and the approach of
winter forced people to return to damaged houses on Santorini
Island, ignoring government recommendations. The government, in turn, decided to demolish most damaged 17th century
mansions and other traditional buildings so as to reduce the
risk of a high death toll.
Whereas by definition the time lag between a short-term
prediction-hours to weeks-and the predicted earthquake is
adequate to undertake emergency measures, for example
evacuation, protection of precious artefacts, etc., it is too short
to permit major work to be undertaken, for example the
reinforcement of structures, and thereby to avoid either the
primary effects (destruction of structures) or the secondary
effects (conflagration due to the failure of a gas network,
epidemics due to contamination after the failure of water pipes,
etc.) of failure. In the case of larger-scale securing operations
(for example in museums, in historical parts of towns with
narrow streets and old buildings, architectural parts of which
may be ready to fall and kill people), the lead time may be
insufficient. It is often accepted implicitly and a priori that
once a short-term prediction of an imminent earthquake is
issued, government agencies and ordinary citizens will respond
rationally to eliminate or reduce the effects of the impending
earthquake. Unfortunately, this is probably overly optimistic.
In contrast, longer-term emergency plans based on assessment
and appreciation of seismic risk factors can reduce or eliminate
most of the causes of damage and fatalities regardless of
whether or not short-term earthquake prediction is possible.
The cost of an evacuation or a period of heightened
preparedness before a predicted earthquake can be extremely
high, especially for regions where tourism is a major industry,
such as many Greek islands, or for institutions such as
museums. In addition to direct losses, a prediction may
482
S. C. Stivos
also have important indirect impact. Rumours of earthquakes
have forced Greek domestic insurance companies to raise
their rates and transfer 98.5 per cent of their risk to foreign
reinsurers (N. Aliferis, Astir Insurance Company, personal
commmunication).
4
SEISMIC SEQUENCES
Predictions of ‘an earthquake’ within a period of a few
weeks would be appropriate for a seismic sequence consisting
of a single main strong and destructive shock (for which a
prediction would be of value) followed by, and possibly
preceded by, a number of minor, harmless shocks (for which a
prediction would have no practical value). However, seismic
sequences in Greece are quite variable and have complicated
patterns. First, most seismic sequences consist of a main shock
and a number of relatively strong aftershocks, or even foreshocks. In several cases, aftershocks have proved to be as
destructive as, or even more destructive than, the main shock
(see Papazachos & Papazachou 1989 and below). Second, even
within a single focal area, both single and multiple main-event
seismic sequences occur, with lag times between these events
usually ranging from minutes to days (Tables 2 and 3). For
example, the 1928 April 22 (M=6.3) and 1930 April 17
( M = 6.0) earthquakes in Corinth were single-event sequences,
but the next sequence, that of 1981, was a triple-event sequence
(Table 3). The Aegean coast north of Volos was hit in 1905 and
in 1911 by single-shock earthquakes; in 1930, a twin seismic
sequence occurred (1905 January 20, M = 6.3,1911 October 22,
M=6.0, 1930 February 23, M=6.0, 1930 March 31, M=6.1;
Papazachos & Papazachou 1989). Hence, as first noticed by
Galanopoulos (1981), and in contrast to most other regions,
when a strong or relatively strong shock hits an area (with the
exception of most intermediate-depth shocks in the southern
Aegean, see Papazachos & Papazachou 1989), the possibility of
another strong earthquake following is not reduced (see, for
instance, the cases of the 1930 Volos and 1981/82 Lesvos
seismic sequences in Table 3). Third, certain areas are hit
frequently (every few tens of years) by seismic sequences with
varying magnitudes (for instance, Rhodes experiences earthquakes with magnitudes ranging between 5.0 and 8.0), so it
is not possible to determine whether a particular event is a
foreshock or a main shock.
Ideally, such complications in the pattern of seismic
sequences would be taken into account by earthquake predictors, yet most predictions issued so far by the various groups
anticipate a single-event seismic sequence (‘an earthquake’)
that will occur within a period of a few weeks. Therefore,
they lack the necessary resolution to recognize whether any
predicted earthquake corresponds to a single event, or to a
multiple-event seismic sequence. Hence, even if such pre-
Table 3. Examples of
sequences in Greece
1957 March 08
1957 April 24/25/26
1980 July 09
1983 January 17
M = 6 . 5 6.8 6.0 in 11 hr
M=6.8 7.2 6.1 in 35 hr
M , = 5.4 6.5 6.1 in 30 min
M = 7 . 0 6.4 in 12 hr
Volos
Rhodes
Volos
Cephalonia
Sources: Papazachos & Papazachou (1989); Makropoulos & Burton
(1981).
long
( 2 4 8 hr) seismic
1881 April 03
1881 April 11
A4 = 6.4
6.0
1894 April 20
1894 April 27
6.4
6.7
Atalandi (Locris)
1930 February 23
1930 March 3 1
6.0
6.0
Volos area
1953 August 09
1953 August 11
1953 August 12
1953 August 12
6.4
6.8
7.2
6.3
Cephalonia (Ionian Sea)
1978 May 23
1978 June 20
1978 July 04
5.8
6.5
5.1
Thessaloniki
1981 February 24
1981 February 25
1981 March 04
6.7
6.4
6.3
Gulf of Corinth
1981 December 19
1981 December 27
1982 January 18
1982January18
7.2
6.5
7.0
5.6
Lesvos island (Aegean Sea)
1986 September 13
1986 September 15
6.0
5.4
Kalamata (SW Greece mainland)
1996 July 26
1996 August 06
1996 November 14
5.2
5.6
5.0
Konitsa (NW Greece)
Chios Island (Aegean Sea); part of
the 1879-1 884 seismic sequence,
which claimed 4000 victims
Sources: Papazachos & Papazachou (1989); Makropoulos & Burton
(1981); Theodoulidis et al. (1996).
dictions were to prove successful, they fail to answer one of the
most pressing questions posed by society: ‘will the earthquake
that has just occurred be followed by another strong, or even
stronger, event?’Arguably, the answer to this question may be
more important than the prediction of the first main shock, as
the second or following events may prove more harmful than
the first one (for example the 1881 Chios and the 1956 Amorgos
earthquakes, Papazachos & Papazachou 1989). The reason is
that buildings are weakened by the first shock and their
resistance to earthquakes is reduced, while certain of their
elements (roof tiles, decorations, broken glass, etc.) are ready
to fall. Under such circumstances, the value of any successful
earthquake prediction is greatly reduced.
5
Table 2. Examples of multiple event, short ( 1 2 days) seismic
sequences in Greece.
multiple-event,
PREDICTED MEIZOSEISMAL AREAS
Burton (1996) discussed the effect of earthquake rupture length
on the evaluation of VAN’S predictions. However, from a
purely utilitarian standpoint, prediction of the meizoseismal
zone of a future earthquake (that is of the area where the
highest seismic intensities will be observed) is of greater
importance. The meizoseismal areas of moderate earthquakes
can usually be modelled as ellipses with a long axis up to a few
tens of kilometres in length (Ambraseys & Jackson 1990); in
such cases, an accurate prediction of the epicentre may well
0 1997 RAS, GJI 131,478434
Earthquake prediction studies in Greece
satisfy the needs of society. By contrast, the meizoseismal areas
of the largest ( M > 7.0) and hence most destructive earthquakes in Greece are irregular. Areas with intensities higher
than VII, separated by distances greater than 600 km, have
been reported by Sieberg (1932; the 1856 Crete and 1926
Rhodes earthquakes of magnitude 8.0-8.2, see Wyss & Baer
1981). Other examples include the 1829 eastern Macedonia
and 1953 Ionian islands earthquakes of magnitude 7.2-7.3
(Papazachos & Papazachou 1989). In such cases, even the most
accurate predictions of epicentral locations will fail to meet the
practical needs of society.
6
COST-EFFECTIVENESS
In Greece, a single earthquake prediction project (VAN)
has absorbed the largest individual share of funds for earthquake research and protection for 15 years (Table 1). Whether
or not the associated predictions were scientifically successful,
they proved of no value to Greek society for two reasons. They
have not, to date, demonstrably contributed to saving lives
or reducing property damage, and they have not contributed
to any emergency planning for future earthquakes. In this
respect, no direct estimation of their social effectiveness can be
made.
It is this author’s view that if all or part of these funds had
been allocated to other fields of earthquake-related research,
much more progress could have be made. Taking, for example,
the area of seismic hazard assessment, the record of earthquakes which are now known to have occurred in Western
Macedonia-an area previously assumed to be ‘aseismic’during the last 2300 years (Stiros 1995b) was compiled with
relatively little effort, and indicates the rate of progress that
may be anticipated, even with relatively limited funding. If the
money spent on earthquake prediction had been used for street
or highway safety (where a death toll of more than 2300 is
recorded every year), more lives than those eventually killed by
earthquakes (less than 10 people per year on average) could
probably have been saved!
In fact, although there are certainly some exceptions,
earthquakes in Greece generally cause relatively few fatalities.
The 1881 Chios earthquakes, for instance, killed about
4000 people. The main reason for this high death toll was
that, after the first major tremor, people were trapped in
narrow streets and were killed by the collapse of weakened
structures during subsequent shocks. Since 1954, more than 50
destructive earthquakes have affected Greece and killed about
250 people in total; of these, eight earthquakes killed more
than 10 people each. In four of these cases (1954 Sophades
earthquake, 54 dead; 1956 Amorgos earthquake, 53 dead; 1965
Arcadia earthquake, 18 dead; 1967 Ag. Eustratios earthquake,
18 dead) most fatalities were due to the collapse of old-style
houses. Such structures are now very uncommon because
building styles and practices in Greece have changed radically.
Estimates of potential future morbidity should therefore be
based on the experience of recent decades.
The death toll in recent decades is mainly due to the
collapse of a small number of multistorey buildings, while
the contribution of other causes was minor. This was the case
in the 1978 Thessaloniki earthquake (about 40 dead), the 1986
Kalamata earthquake (18 dead) and the 1995 Aigion earthquake (25 dead). This death toll could be much more effectively
eliminated by structural intervention with respect to weak
01997 U S , GJI 131,478484
483
buildings than by predicting earthquakes, even if prediction
were scientifically feasible.
In many cases, it is relatively easy to identify vulnerable
buildings, especially those in which large gatherings are
regularly held (e.g. schools, churches, hotels, etc.). During the
last six months one prefabricated school in Athens suddenly
collapsed, while a second school was abandoned when, during
a routine examination, it was discovered that iron in reinforced
concrete columns was dramatically corroded. Buildings next
to rocks ready to fall, or old buildings in which important
changes have recently been made (for instance, where columns
or beams have been removed, or large openings in load-bearing
walls have been made after a change of use from houses to
shops), can be easily identified. Part of the cost of intervention
can be recovered from insurance expenses and other sources
(funds for restoration of traditional buildings, etc.).
It therefore appears likely that the benefits of any possible,
successful short-term earthquake prediction, such as those
typically announced in the last 15 years, are drastically smaller
than has been a priori assumed. Furthermore, most of the
anticipated benefits could be achieved with greater reliability
through longer-term preparedness measures. Hence, at least
at the present stage, earthquake prediction is certainly not
cost-effective in Greece.
ACKNOWLEDGMENTS
I acknowledge with thanks the benefit of discussions with and
material sent by R. J. Geller and K. Thanassoulas, as well
as copious reviews by R. J. Geller, J. R. Evans and two
anonymous reviewers.
REFERENCES
Ambraseys, N.N. & Jackson, J., 1990. Seismicity and associated strain
of central Greece between 1890 and 1988, Geophys. J. Int., 101,
663-708.
Anagnostopoulos, S., 1995. VAN, funds and antiseismic policy,
Information Bull. Technical Chamber of Greece, 1872,ll September,
8 4 8 6 (in Greek).
Asteriadis, G. & Livieratos, E., 1989. Pre-seismic responses of underground water level and temperature concerning a 4.8 magnitude
earthquake in Greece on October 20, 1988, Tectonophysics, 170,
165-169.
Asteriadis, G. & Zioutas, G., 1990. Application of an outlier
detecting method for identifying changes in the underground
water level caused by crustal movements, Geophys. J. Znt., 103,
415424.
Bernard, P., 1992. Plausibility of long distance electrotelluric precursors to earthquakes, J. geophys. R e x , 97, 17 531-17 546.
Bernard, P., Pinettes, P., Hatzidimitrou, P.M., Scordilis, E.M., Veis, G,
& Milas, P., 1997. From precursors to prediction: a few recent cases
from Greece, Geophys. J. Int., 131, 467-477 (this issue).
Bornovas, J. & Rondogianni, T. 1983. Geological Map of Greece,
1:500 000 scale, IGME, Athens.
Burton, P., 1985. Electrical earthquake prediction, Nature, 315,
370-371.
Burton, P., 1996. Dicing with earthquakes, Geophys. Rex Lett., 23,
1379-1382.
Constantinides, D., 1987. Preseismic change of underground water
level. Observations for a prediction of an earthquake from the
fluctuation of the level of acquifers of Thessaly, Bull. geol. SOC.
Greece, 19, 219-232 (in Greek).
484
S. C. Stiros
Drakopoulos, J & Latoussakis, .I.,1987. Application of statistical
criteria for identification of preseismic sequences and prediction
of major main earthquakes in Greece, Bull. geol. Soc. Greece, 19,
463-475 (in Greek).
Drakopoulos, J. & Stavrakakis, G.N., 1996. A false alarm based on
electrical activity recorded at a VAN station in northern Greece in
December 1990, Geophys. Res. Lett., 23, 135551358,
Galanopoulos, A,, 1981. The damaging earthquakes and the earthquake potential of Greece, Ann. Geol. Pays Helleniques, 30,
647-724 (in Greek).
Geller, R.J., 1996. VAN: A critical evaluation, in A Critical Revieu of
VAN, pp. 155-238, ed. Lighthill, J., World Scientific, Singapore.
Gruszow, S., Rossignol, J.C., Tzanis, A. & Le Mouel, J.L., 1996.
Identification and analysis of electromagnetic signals: the case of the
Kozani VAN prediction, Geophys. Res. Lett., 23,2025-2028.
Ifantis, A,, Tselentis, G.-A,, Varotsos, P. & Thanasssoulas, C.,
1993. Long-term variations of the Earth’s electric field preceding
two earthquakes in Greece, Acta Geophysica Polonica, 41,
337-350.
IGRS (Institut de Geologie et des Recherches du Sous-Sol) & IFP
(Institut Francais du Petrol), 1966. Etude Geologique de I’Epire,
Technip, Paris.
Jackson, J., Gagnepain, J., Houseman, G.A., King, G.P.,
Papadimitriou, P., Soufleris, C. & Virieux, J., 1982. Seismicity,
normal faulting and the geomorphological development of the Gulf
of Corinth (Greece). The Corinth earthquakes of February and
March 1981, Earth planet. Sci. Left., 57, 371-397.
Kagan, Y. & Jackson, D., 1996. Statistical tests of VAN earthquake
predictions: comments and reflections, Geophys. Res. Lett., 23,
1433-1436.
Labeyrie, J., 1993. Comments on the results obtained in Greece
by the VAN method of seismic prediction, Tectonophysics, 224,
161-167.
Makropoulos, K. & Burton, P., 1981. A catalogue of seismicity in
Greece and adjacent areas, Geophys. J. Int., 65,741-762.
Masood, E., 1995. Court charges open split in Greek earthquake
experts, Nature, 377, 315.
Mulargia, F. & Gasparini, P., 1992. Evaluating the statistical validity
beyond chance of ‘VAN’ earthquake precursors, Geophys. J Int.,
111,3244.
Papadopoulos, G., 1988a. A note on the prediction of the September
13, 1986, strong earthquake in Kalamata, Tectonophysics, 145,
337-341.
Papadopoulos, G., 1988b. Long-term accelarating foreshock activity
may indicate the occurrence of a strong shock in the western
Hellenic arc, Tectonophysics, 152, 179-192.
Papadopoulos, G. 1996. On the successful short-term earthquake prediction of the 15 June 1995 Greek earthquake (Ms=6.0) from an
earthquake synchronization process, Abstracts, Int. Mtg results of
the May 13, 1995 earthquake of West Macedonia: one year after,
Kozani, Greece, 24-28 May.
Papanikolaou, D., 1993. The effect of geological anisotropies on the
detetability of seismic electric signals, Tectonophysics, 224,181-187.
Papazachos, B. & Papazachou, C., 1989. Earthquakes in Greece, Zitis,
Thessaloniki.
Papazachos, B., Karakaisis, G., Karakostas, B., Papadimitriou, E.,
Chadjidimitriou, P., Komninakis, P. & Tassos, S., 1987.
Contribution to the study of anomalous changes of geoelectric field
in the effort of predicting earthquakes, Bull. geol. Soc. Greece, 19,
401430 (in Greek).
Sieberg, A. 1932. Untersuchungen ueber Erdbeben und Bruchschollenbau im oestlichen Mittelmeersgebiet, Denkschrijten der
med.-natur. w. Ges. zu Jena, 18, Band 2.
Stavrakakis, G. & Drakopoulos, J., 1996. The VAN method:
contradictory and misleading results since 1981, Geophys. Res.
Lett., 23, 1347-1350.
Stavrakakis, G., Drakopoulos, J. & Latousakis, J., 1990. Evaluation of
the correlation of SES with earthquakes in the area of Greece: time
period 1987-1989, Publ. Geodyn. Inst. National Observatory of
Athens, 1.
Stiros, S., 1994. Anomalous attenuation of seismic waves in the
lonian Islands, Greece, EOS, Trans. Am. geophys. Un., 75,
117-118.
Stiros, S., 1995a. Archaeological evidence of antiseismic constructions
in antiquity, Ann. Geofis.. 28,725-736.
Stiros, S., 1995b. Unexpected shock rocks an ‘aseismic’ area, EOS,
Trans. Am. geophys. Un.. 76,513 & 519.
Stiros, S., 1997. Historical seismicity, palaeoseismicity and
seismic risk in western Macedonia, northern Greece, J. Geodyn.,
in press.
Stiros, S., Pirazzoli, P., Laborel, J. & Laborel-Deguen, F., 1994.
The 1953 earthquake in Cephalonia (western Hellenic arc):
coastal uplift and halotectonic faulting. Geophys. J. Int., 117,
834-849.
Thanassoulas, K., 1984. Application of the geoelectric method in the
study of salt deposits in the Monolithi area of Ioannina, IGME
unpublished Report, 4184.
Thanassoulas, C., 1991. Determination of the epicentral area of
three earthquakes (Ms>6) in Greece, based on electrotelluric
currents recorded by the VAN network, Acta Geophysica Polonica,
39,373-387.
Thanassoulas, C. & Tselentis, G., 1993. Periodic variations in the
Earth’s electric field as earthquake precursors: results from recent
experiments in Greece, Tectonophysics, 224, 103-1 11.
Theodoulides, N., Papaioannou, C., Dimosthenous, M. &
Demetriou, P., 1996. The earthquake of Konitsa of August 6th,
1996 (Ms=5.6), Information Bull. Technical Chamber of Greece,
1919,22-32.
Underhill, J., 1989. Late Cenozoic deformation of the Hellenide foreland, western Greece, Geol. Soc. Am. Bull., 101, 269-282.
Utada, H., 1993. On the physical background of the VAN earthquake
prediction method, Tectonophysics, 224, 153-160.
Varotsos, P., 1987. Comment on ‘Correlation of electrotelluric signals
with earthquakes preceding and following earthquakes’, Bull. geol.
Soc. Greece, 19,499-500.
Varotsos, P. & Alexopoulos, K., 1984. Physical properties of the
variations of the electric field of the earth preceding earthquakes, 1,
Tectonophysics, 110,73-98.
Varotsos, P. & Lazaridou, M., 1991. Latest aspects of earthquake prediction in Greece based on seismic electric signals, Tectonophysics,
188,321-347.
Varotsos, P., Alexopoulos, K. & Nomikos, K., 1981. Seven-hour precursor to earthquakes determined from telluric currents, Pruktika
Akadimias Athinon, 56,417433 (in Greek).
Varotsos, P., Alexopoulos, K. & Lazaridou, M., 1993. Latest aspects
of earthquake prediction in Greece, based on seismic electric
signals, 11, Tectonophysics. 224, 1-37.
Varotsos, P., Eftaxias, K.,Vallianatos, F. & Lazadidou, M., 1996. Basic
principles for evaluating an earthquake prediction method,
Geophys. Res. Lett., 23, 1295-1298.
Wyss, M. & Baer, M., 1981. Earthquake hazard in the Hellenic arc,
in Earthquake Prediction, an International Review, pp. 153-172,
eds Simpson, D. & Richards, P., Maurice Ewing Series 4, AGU,
Washington, DC.
Yoshii, H., 1993. Social impacts of earthquake prediction in Greece,
Tectonophysics, 224,251-255.
0 1997 RAS, GJI 131,478484