Broadband stations in Western Corinth Gulf
J. Zahradnik1 and G.-A. Tselentis2
Corinth Gulf separates the continental Greece from Peloponnese peninsula.
It represents an extensional tectonic structure, belonging to the seismically
most active regions of the Mediterranean. The last destructive event of M=6.2
occurred in the Gulf close to the city of Egio in June 1995. The city of Patras,
hub of the western Peloponnese, was damaged by the 1993 earthquake.
The key role in understanding those earthquakes, and predicting ground motions
during similar future events, is played by detailed seismic observations.
Western part of the Corinth Gulf has been monitored since 80's by PATNET,
the short-period telemetered seismological network of the Patras University,
covering the whole western Greece. Since November 1997, three broadband stations
operate in the Gulf as a long-term temporary network, jointly managed by the
Charles University and Patras University.
Station coordinates and data
On November 18, 1997, one of the seismometers was being deployed just during
strong shaking due to M=6.6 earthquake in Zakynthos region, six minutes later
followed by an aftershock M=5.9 (epicentral distance of about 100 km). During
the following night, hundreds of aftershocks were already recorded at two of
these BB stations.
Fig. 1. Broadband (red) and PATNET (green) stations, shown together with
EGIO record of the Zakynthos earthquake, made during first minutes of
the station 'life'.
Since July 1998, the recording sites have remained as follows (Fig. 1):
CLAU station at Achaia Clauss, close to the city of Patras, EGIO at Kumari
village, close to the city of Egio, and KALI at Sergoula village.
The first two are situated on the southern coast of the Gulf.
KALI is on the northern coast, accessible from Patras by the Rio-Antirio ferryboat.
CLAU seismograph lies on the cemented floor of a small brick house adjacent to a
church, close to famous wine yards and wine factory. EGIO seismograph is on a
concrete block forming part of the basement and one wall of a masonry uninhabited
farmer house. KALI instrument is on a tile-covered ground floor of an unused
village school building. All the three are powered from 220V mains, and switch
automatically to rechargeable 12V batteries in case of a power brake.
The northern coast exposes bedrock in many places, while the southern cost is
mostly covered by sediments. KALI is on limestone, in a steeply sloping coastal
mountain range, at the altitude of about 400 m above the sea level, with a
magnificent view of the Gulf (thus KALI name, meaning 'good', or 'nice' in Greek).
CLAU and EGIO are on the less compact sedimentary rocks, both in a hilly terrain
at altitudes of about 200 m.
All EGIO records show notable site effects (e.g. in Fig. 2), such as long
duration, and frequency-selective amplification.
Fig. 2. Example of the BB records for a local event belonging to a cluster
between the KALI and EGIO station (Fig. 4). EGIO record (top three traces)
strongly differs from KALI (bottom), although epicentral distance is the same.
CLAU (midlle) has the weakest motion, but EW is disturbed by a 'mouse'.
All three stations are completely equipped by the Guralp hardware and software.
As such, they probably belong to few stations of the world where seismic products
of different companies are not mixed with each other. Each station is composed of
a CMG-3T three-component seismometer, DM-24 digitizer, GPS antenna, and SAM data
logger with a 3.5MB RAM buffer, and 2GB SCSI hard disk. Recording is in a very
efficient binary Guralp Compressed Format, GCF, sometimes also called SAM.
The records clip (5,500,000 counts) at velocities around 2 millimeters per second.
This happens, for example, at magnitudes above 4 from epicentral distances of a
few kilometers. That is why CMG5-T strong-motion accelerographs will accompany
the BB stations, starting in 1999. The minimum noise level (at KALI and EGIO)
is around 500 counts; in CLAU it is usually higher because of proximity of the
wine factory. Oceanic microseisms and wind are highly variable, sometimes up to
20,000 counts during stormy days.
Three velocity components (N,E,Z) are recorded, with their mutual orthogonality
guaranteed by the fact that all sensors are housed in the same box, while the
N-S orientation of the box is only approximately determined by compass.
The continuous channels have sampling rate of 20 Hz. The triggered recording is
controlled by the Z-component, band-pass filtered for the trigger algorithm
between 5 and 45 Hz, and it has sampling rate of 200 Hz.
Several combinations of the triggering parameters were tested, which finally
settled at STA = 1 s, LTA = 50 s, STA/LTA = 6, pre-trigger time = 40 s,
post-trigger time = 70 s. This gives about 100 triggers at each station per
month during normal activity in the Gulf, on average.
Besides the ground-motion velocity, also the mass position (N,E,Z) is
continuously recorded at 4 Hz sampling. Information about the GPS performance,
and SCSI disk activity is recorded as text files.
As a rule, all streams together represent less than 15 MB per day. Thus a
2GB disk would fill roughly in 4 months. However, because the stations are
fully unattended, they require more frequent visits.
Most technical problems of the first twelve months were connected with the
digitizer. These were solved by Guralp engineers on-site, and by sending one
station for repair to England.
Fig. 3. Step in the mass channel at CLAU station, EW component, corresponding
to the 'mouse' of Fig. 2.
The most unpleasant remaining problem is 'mice' on the horizontal components.
By 'mouse' we mean a spurious velocity wave, about 1 minute long resembling a
long-period calibration impulse (Fig. 2). This is accompanied by a step motion
on the corresponding mass channel (Fig. 3). The 'mice' occur during onsets of
some local seismic events. At a given station 'mice' are more probable for
larger ground motions (still below the clipping level).
As a rule, one horizontal component at each station suffers more than the other,
independently of the event location. For example, EW component at CLAU is
always more vulnerable than NS. Moreover, the EW component at CLAU has a 'mouse'
already at ground-motion levels at which the other stations remain undisturbed.
The 'mice' are not clearly dependent on the centering state. They may occur
already at a very well centered state, e.g. at less than one third of the full
mass position range. Our primitive interpretation is that the force-balance
system does not fully compensate the mass movement, thus it 'jumps', and
system responds in a way similar to the step calibration. We think that
'mice' are a challenge for the instrumental specialists.
The 'mice' represent a strong limitation of the source inversions.
In particular, the important low-frequency part of spectra (f < 0.1 Hz),
well above the noise level for nearby local events of magnitudes around 3,
It is to mention that, besides the visible 'mice', there is also a less
obvious instability of the horizontal components on practically any onset
of a local event. As a result, frequencies below 0.1 Hz would need a
(non-standard) instrumental correction during data processing, in general.
The stations are visited every 3 months. In an ideal case only centering
is needed. Every visit includes the SCSI disk exchange, and its replay
onto PC hard disk at the Patras University, followed by copying all streams
as a whole on CD's. Within 10 days, local events recorded at all three stations
with a good S/N ratio are selected in Prague, and extracted from CD's.
Binary GCF data, and data converted into PC-SUDS format are exposed
on an anonymous ftp server. The Patras group performs the arrival-time
reading and HYPO location (including the PATNET data). This, in fact,
is already a re-location, since Patras network locates significant earthquakes
continuously. Next processing of the BB records of the relocated events is
carried out again in Prague. This comprises both analyses of the binary
data (user-friendly Guralp software Scream and GCFInfo), as well as conversions
Current processing of the ASCII data includes the baseline correction and
trend removal, filtration, rotation, spectra, focal mechanisms from spectra
and polarities, and modeling waveforms by synthetic seismograms
(home made FORTRAN codes).
Fig. 4. Epicenters of the selected events.
Note a cluster between KALI and EGIO stations.
Data. The local broadband waveform data can be downloaded from
the server at Prague.
The data are organized chronologically, according to the station visits
(e.g., cd_4, 5, 6, etc.).
Location. Advanced re-location methods, such as differential evolution and fully
non-linear hypocenter determination have been developed and applied to the
local events (Kvasnicka and Ruzek, 1999; see
The focal-mechanism inversion, based on the amplitude spectra of the BB
records (0.2 to 1 Hz) and PATNET first-motion polarities, has been developed
and applied to the clustered events between KALI and EGIO (Zahradnik, 1999;
see web site).
The site effect at EGIO has been investigated by means of spectral ratios
EGIO/KALI from the teleseismic records (Jansky, 1999; see
web site 1 and
web site 2).
Of course, the main usage of the regional and teleseismic records is expected
in a broader international co-operation. Therefore, since 1999, the Charles
University has joined Orfeus, and the continuous streams will be copied on
CD's and provided every 3 months.
N. Melis (EPPO, Athens) provided a very useful advice. E. Sokos and A. Sotiriou
(Patras Univ.) assisted in the station deployment. M. McGowan and
Ch. Pearcey (Guralp Co.) provided the software and instrumental help.
K. Papatsimpa (Patras Univ.) processed the PATNET records.
J. Jansky, M. Kvasnicka, V. Plicka and I. Oprsal (Charles Univ.)
participated in the fieldwork and data processing. The financial support
from two EU Inco-Copernicus projects, COME and ISMOD, and the Charles
University grant 5/97/B is appreciated.
Two broadband stations were purchased by the Faculty of Mathematics
and Physics, Charles University.