The two destructive earthquakes of 1995 in Greece, the May 13 Ms=6.6 K
ozani-Grevena and the June 15 Ms=6.2 Aigion events, provide interestin
g material for analysing problems related to the identification of pre
cursors and to the efficiency and usefulness of prediction. The Kozani
earthquake was preceded, within 30 minutes of the main shock, by five
foreshocks with magnitude greater than 3.5 (Papazachos et al. 1995).
We relocated these events with respect to each other, showing that the
y are clustered within 2 km of one another, about 5 to 10 km to the SS
W of the main-shock epicentre. This size of foreshock clustering corre
ctly fits the correlation law with the main-shock magnitude obtained b
y Dodge, Beroza & Ellsworth (1996) for Californian earthquakes. These
foreshocks led to people leaving their houses, which explains the abse
nce of casualties, despite the partial destruction of several villages
. The possibility of issuing predictions in this area from the observa
tion of earthquake clustering is analysed in light of the seismicity o
bserved during the last 15 years. A prediction was issued by the VAN g
roup before this earthquake, based on SES signals (IOA station, 18-19
April 1995), which is considered by VAN as a success (Varotsos et al.
1996a), but is in fact a failure to predict (Geller 1996). This SES wa
s also recorded by a magnetotelluric station installed by IPGP, a few
kilometres from IOA (Gruszow et al. 1996). Gruszow et al. (1996) sugge
sted an artificial origin for the SES, but could not track it. Simple
amplitude estimates show that a local, natural source such as an elect
rokinetic effect is unlikely, and that a remote electrokinetic source
in the epicentral area can be even more confidently rejected. Another
SES on VAN's network (VOL station, 30 April 1995) led the VAN group to
predict an earthquake outside the IOA sensitivity area (IOA did not r
ecord any anomaly), and to announce a success when the Aigion earthqua
ke occurred (Varotsos et al. 1996a); however, this event was located i
nside the IOA sensitivity area, and the prediction was hence a failure
(Wyss 1996; Geller 1996; Bernard et al. 1997). Furthermore, at the ti
me of this SES, no tilt nor strain was observed above the noise level
of a few 10(-8) at the IPGP/NTUA Galaxidi geophysical observatory, 20
km from the hypocentre, leading Pinettes et al. (1996) to conclude tha
t the electrical source of this SES was most probably located near VOL
, 100 km away, whatever its correlation with the earthquake.