Rapid Source Parameter Determination of the
In this note, I report about the source parameter determination of the Izmit
eartquake as it was done at GFZ in the morning of August 17, 1999.
The results were distributed via electronic mail to customers of the
European-Mediterranean Seismological Centre (EMSC). Two messages were
disseminated. The first one at 08:08 UTC was based only on data from
5 stations, for the second one posted at 10:17 UTC, about 10 hours
after the event, data from 12 stations could be used. In the
following I shall briefly describe the procedure, the problems that
arose on this day and present seismogram examples that illustrate
various parameters of the solution.
August 17, 1999, Izmit Earthquake at GFZ Potsdam
Fig. 1. Distribution of stations available in GEOFON Spyder® online data
pool on the morning after the August 17, 1999, Izmit eartquake whose
epicentral location is depicted by the red star.
The rapid source parameter determination at GFZ makes use of waveforms
that are made availble in the global Spyder®, usually
within a few hours after the event. Within Europe Spyder® data is available
Automatic data retrieval by the Spyder® system is
triggered by alarm messages that are received from
As part of testing a regional Spyder® system set up for eartquakes in
the European-Mediterranean area, several alarm messages of the Turkey
event were received from EMSC in addition to the one of NEIC. As a
result, the Spyder® retrieval system got stalled and had to be
restarted manually in the morning of August 17. So we faced the
problem in the morning of August 17 that only relatively few data had
become available for a source parameter inversion, and most of these
data were from stations at regional distances. Most data from
teleseismic distances (D ³ 30°) started to flow into the online
datapool in the afternoon beginning at 17:31 UTC.
The stations for which data had become available are depicted in Fig. 1.
Most of the stations lie in the European area. Stations that were
not suitable for the source mechanism inversion are plotted as red
dots, however, some of these stations had useful vertical-component
waveforms that were used to constrain focal depth and/or source
As a first step in the inversion an attempt is made to constrain focal
depth and source duration. This is done by visual inspection of
broadband waveforms. Examples are shown in Figs 2-3.
2. Comparison of observed P wave displacement seismogram (blue) observed at WET
(D = 14.76°) with synthetics for a source duration of 20 s and a variety
of focal depths. The red synthetic is for h=15 km.
Shown in Fig. 2
is the vertical-component displacement seismogram of the P wave
recorded at the GRSN station Wettzell (WET). It illustrates the way
of how focal depth is constrained by interactive interpretation of
depth phases. The interpretation is shown by the picks of P and pP
phases in the bottom trace. Comparison of the observed blue waveform
with synthetics calculated with the reflectivity method for a range
of focal depths is shown in the middle and upper part of Fig. 2. The
red trace is the synthetic for a focal depth of 15 km which was later
adopted in the source mechanism inversion. The evaluation of depth
phases for all stations gave a mean depth of 17 km.
Fig. 3 illustrates the way of how an estimate of source duration
is obtained. This again is done interactively on P wave displacement
seismograms. The duration of the P wave pulse is taken as a measure
for the source duration. We use the formula given by Brüstle and
Müller (1983) which approximates the moment release of a point
source by a half-sinusoid from 0 to its final value M0. The far-field
displacement is proportional to the time derivative of this
source-time function which has the form of a half cosine function.
This is a very simple model of a source-time function which does not
account for any complexities that may arise from multiple sources.
Also, directivity effects are not considered. The approach is very
subjective, but the interpreter can be supported by displaying
reflectivity synthetics for a variety of source durations and
comparing them with the observed seismogram as shown in Fig. 3. In
the example shown the measured P wave duration was 16.5 s; the
average from all observations was 18 s.
Fig. 3. Comparison of P wave displacement seismogram recorded at PAB (San
Pablo, Spain, D = 26.14°)
with synthetics calculated for 10 km focal depth and a variety of
source durations. The »20 sec« synthetic (in red) is a
good approximation for the Izmit earthquake.
A non-linear grid search algorithm is used which minimizes the
difference between observed and theoretical P/S amplitude ratios. The
method has been described at greater detail by Bock (1993) and Bock
et al. (1994). It resembles in many ways the relative amplitude
method described by Pearce (1977).
A brief account of the method can be found at the
GFZ web site.
The peak-to-peak amplitude of the
P wave is measured on the vertical component and that of the S wave
on the vertical, radial- and tangential-horizontal components.
Amplitudes are measured over a full wavelength so that the estimate
may contain besides P also pP and sP in case of shallow events. The
time windows over which peak-to-peak P and S wave amplitudes were
measured are indicated for station PAB in Fig. 4.
Fig. 4. Example of picking P and S wave amplitudes.
The P/S amplitude ratios are the entry parameters for the grid
search algorithm. Observed amplitude ratios are compared with
synthetic ratios obtained with the reflectivity method.
Reflectivity synthetics are stored for distances up to 80°
and 5 km steps in focal depth. This explains the fact that focal
depths adopted in the inversion are rounded to the nearest 5 km
interval. First, a rough search is conducted in 10°
interval for strike, dip and slip of a double couple source. For the
Izmit earthquake and other events in the EMSC area the results of the
inversion are published at a
GFZ web page. This page
displays a table of events with links to the files containing the
detailed description of source parameters. For reasons outlined at
the beginning of this article, the list of stations used in the
source mechanism inversion contains only stations at regional
distances. This may explain the relatively large value for the moment
magnitude as compared to Harvard (7.5) and USGS (7.4). It also
illustrates the need to compare observed amplitude ratios with
reflectivity synthetics to account for wave propagation through the
Observed and synthetic waveforms are compared in Fig. 5. The
station distribution is far from ideal as data are from one quadrant
in azimuth only. Despite the fact that there are discrepancies
between synthetics and observations, the overall P/S amplitude ratios
are well matched by the proposed solution. The overall error of the
focal mechanism is estimated to be about 15°
for strike, dip and rake based on the distribution of the misfit
function. Later modelling of broadband waveforms using the dataset
distributed by IRIS provided better constraints on the source-time
function and spatial extent of the Izmit event (Bock et al., 1999).
In particular, we believe that the main event was followed by two
more subevents to the east of rupture onset. These two events show up
in Fig. 3 at about 30 s and 42 s relative to the centroid time of the
Fig. 5. The EMSC source mechanism of the August 17, 1999, Izmit
eartquake (bottom), and comparison of observed with synthetic
waveforms. The seismograms were filtered with a 3-pole Butterworth
bandpass with corners at 0.02 and 0.1 Hz.
Bock, G., 1993. The Woods Reef (New South Wales) earthquake of 14
November 1990: Focal mechanism derived from amplitude ratios and
synthetic seismograms, Australian Journal of Earth Sciences,
Bock, G., Hanka, W. and Kind, R., 1994. EMSC rapid source
parameter determination, EMSC Newsletter, 6, 2-4.
Bock, G., Tibi, R., Baumbach, M., Grosser, H., Milkereit, C.,
Kind, R. And Zschau, J., 1999. Rupture process of the Great Izmit
(Turkey) earthquake of August 17, 1999, Invited poster, AGU Fall
meeting, December 1999.
Pearce, R. G., 1977. Fault plane solutions using relative
amplitudes of P and pP, Geophys. J. R. Astr. Soc.,