next up previous contents
Home: Berkeley Seismological Laboratory
Next: Finite Source Modeling of Up: Ongoing Research Projects Previous: Parkfield Research

Source Rupture of the Aug. 17, 1999 Izmit/Turkey Earthquake

Ludovic Brégera, Fumiko Tajima, Douglas Dreger, and Barbara Romanowicz


The Aug. 17, 1999 Turkey earthquake (BDSN surface wave magnitude Ms=7.8) took place along a segment of the North Anatolian fault which had been a seismic gap of over 100 km between the two earthquake ruptures of 1963 (M6.3), and 1967 (M7.2) [Toksöz et al., 1979]. The automated Centroid Moment Tensor (CMT) solution determined for this event by the Harvard group shows an almost pure right lateral strike slip event (strike=268o, dip=84o, and slip=180o), with a source depth of 16 km. Three months later another large event (BDSN: Ms=7.4; Harvard CMT solution: strike=265o, dip=65o, and slip=-158o) occurred $\sim $100 km east of the August epicenter. The North Anatolian fault has a rich history of repeating earthquakes and is known to involve progressive failure of adjacent segments. Understanding the triggering mechanism of adjacent segments of fault requires knowledge of the slip distribution of previous events. We conducted a preliminary analysis of the distributed fault slip and an estimate of source rupture process, based on teleseismic observations, that allows varying mechanisms. In the analysis, we used bodywave displacement seismograms recorded by the Incorporated Research Institutions for Seismology (IRIS)/ Global Seismographic Network (GSN) stations at teleseismic distances.

Slip distribution

A kinematic model of the distributed fault rupture was obtained using the method of Antolik [1998], and Antolik et al. [1999]. This method inverts seismic moment rate functions that may be derived by either empirical Green's function deconvolution, from a smaller collocated event with the same radiation characteristics, or by the deconvolution of theoretical Green's function, for an appropriate source radiation pattern and depth. In either case the deconvolution serves to remove path effects, leaving an estimate of the seismic moment rate. For the Izmit earthquake, we obtained the seismic moment rate using theoretical Green's functions computed using the reflectivity method, and the focal parameters obtained by the Harvard group.

The inversion suggests a bilateral rupture extending 50 km to the west and 60 km to the east of the hypocenter with its major slip occurring to the west (Figure 14.1, 14.2). The average and peak slips are found to be 4.2 and 9.3 m, respectively. The rupture propagated upwards, with most of the slip occurring near the surface. The total areas of displacement is estimated to be about 1200 km2.

Figure 14.1: Epicenters of the August 17 and November 12, 1999 earthquakes and previous major events in this region. The focal mechanisms of the 1999 events are based on the Harvard CMT solutions. Yellow dots indicate the aftershock locations since Aug. 17. We also plotted the surface trace of the grid used in the inversion (red line).
\epsfig{, width=7cm}\end{center}\end{figure}

Figure 14.2: Slip distribution obtained by inversion of the P MRFs at stations ADK, INCN, KONO, MA2, PAB, SJG, and YSS. The main slip occurs to the west of the epicenter, but there is evidence for a significant amount of slip to the east, which suggests a bilateral rupture. Yellow dots indicate the aftershock locations during the first 4 months following the main shock. The position of the hypocenter is given by the yellow star.
\epsfig{, width=12cm, height=5cm}\end{center}\end{figure*}

Figure 14.3: Inversion result, i.e., three subevents (E1, E2, and E3), source-time function, and spatial distribution of three point sources in the grid on the given fault plane.
\epsfig{, width=12cm, height=7cm}\end{center}\end{figure*}

Source Rupture Process

We have also employed the inversion method of Kikuchi and Kanamori [1991] that models the entire source process as a series of point source subevents and thus permits the determination of subevent moment-tensors, point-source locations, and the time-function simultaneously.

The main source process can be explained by several subevents (Figure 14.3). The first subevent (E1) is determined with a symmetric trapezoidal unit time function ($\tau_{1}$,$\tau_{2}$)=(3s, 6s) at the NEIC epicenter with a source depth of $\sim $10 km and a released moment 0.96 x 1020 Nm or Mw=7.2). The mechanism of this subevent is almost a pure right lateral strike slip ( strike=275o, dip=89o, and slip=-180o), similar to the CMT solution. Smaller subevents, E2 and E3 (($\tau_{1}$,$\tau_{2}$)=(3s, 3s), and (1.5s, 1.5s)) with a reversed strike slip fault motion and moment release of $\sim $0.3 x 1020 Nm appear to have occurred $\sim $30 (at 12 s) to 60 km (at 34 s) east of the first subevent. However, the determination of these smaller subevents was not stable being affected by a small change of the input parameters such as the unit time functions, weighting factors for the individual waveform data etc... The trade-offs between the source parameters of smaller subevents and structural noise seems to be also significant in this frequency band. If the first subevent corresponds to the major slip determined to the west of the epicenter, then the stress drop of this event is roughly estimated to be 7.6 Mpa according to the following formula:

\begin{displaymath}\Delta\sigma = \frac{2.5M_{o}}{S^{1.5}}\end{displaymath}

where S=50 x 20 km2


Our preliminary analysis using two different approaches suggests a bilateral propagation of the main source rupture. While the two separate portions of the slip distribution were identified on the fault plane, the major slip occurred to the west of the epicenter. The first subevent determined in the source process has a moment release of 0.96 x 1020 Nm and may correspond to the rupture of the main asperity which has a dimension of 50 x 20 km2. Adopting the moment and slip dimension, the stress drop of this event is estimated to be 7.6 Mpa. The slip distribution to the east of the epicenter roughly coincides with the location of the second subevent which was determined $\sim $12 s after the first subevent in the source process. Due to the trade-off between the subevent source parameters and structural noise the determination of the smaller subevents is not stable. On the other hand strong ground motion seismograms recorded at several stations near the source area such as YPT ($\sim $20 km west of the epicenter) indicate two distinct events which occurred with a time interval of $\sim $30 s. The separation of the two events decreases in the waveforms recorded by stations at further distances to the west, which may offer a support for the second subevent to have occurred to the east of the epicenter. The particle motions obtained from the strong motion data at Yarimca (YPT) Station also indicate a bilateral propagation of the rupture.

In general, the determination of earthquake source parameters (moment tensor, hypocenter, and source-time function) by waveform inversion is a non-linear and multi-dimensional problem in which trade-offs between the parameters can be significant and complicated. The source process determination for this event will benefit from a careful analysis of the inversion parameters, but also from better-constrained velocity models in different azimuths of ray path propagations. Preliminary results are nevertheless consistent with other studies, and encouraging.


The authors benefited from discussions with Roland Bürgmann.


Antolik, M., New results from studies of three outstanding problems in local, regional, and global seismology, Ph.D. Dissertation, Univ. Calif. Berkeley, 1996.

Antolik, M., D. Dreger, and B. Romanowicz, Finite fault source study of the great 1994 Bolivia earthquake, Geophys. Res. Lett., 23, 1589, 1996.

Kikuchi, M., and H. Kanamori, Inversion of complex body waves-III, Bull. Seismol. Soc. Am., 81, 2335, 1991. Toksöz, M., A. F. Shakal, and A.I. Michael, Space-time migration of earthquakes along the North Anatolian fault zone and seismic gaps, Pure Appl. Geophys., 117, 1258, 1979.

next up previous contents
Next: Finite Source Modeling of Up: Ongoing Research Projects Previous: Parkfield Research

The Berkeley Seismological Laboratory, 202 McCone Hall, UC Berkeley, Berkeley CA 94720
Questions and comments to
Copyright 2000, The Regents of the University of California.