Analysis of Izmit aftershocks 25 days before the November 12th 1999 Düzce earthquake, Turkey

Abstract

We investigate spatial clustering of 2414 aftershocks along the Izmit M_w = 7.4 August 17, 1999 earthquake rupture zone. 25 days prior to the Düzce earthquake M_w = 7.2 (November 12, 1999), we analyze two spatial clusters, namely Sakarya (SC) and Karadere–Düzce (KDC). We determine the earthquake frequency–magnitude distribution (b-value) for both clusters. We find two high b-value zones in SC and one high b-value zone in KDC which are in agreement with large coseismic surface displacements along the Izmit rupture. The b-values are significantly lower at the eastern end of the Izmit rupture where the Düzce mainshock occurred. These low b-values at depth are correlated with low postseismic slip rate and positive Coloumb stress change along KDC. Since low b-values are hypothesized with high stress levels, we propose that at the depth of the Düzce hypocenter (12.5 km), earthquakes are triggered at higher stresses compared to shallower crustal earthquake. The decrease in b-value from the Karadere segment towards the Düzce Basin supports this low b-value high stress hypothesis at the eastern end of the Izmit rupture. Consequently, we detect three asperity regions which are correlated with high b-value zones along the Izmit rupture. According to aftershock distribution the half of the Düzce fault segment was active before the 12 November 1999 Düzce mainshock. This part is correlated with low b-values which mean high stress concentration in the Düzce Basin. This high density aftershock activity presumably helped to trigger the Düzce event (M_w = 7.2) after the Izmit M_w 7.4 mainshock.

Introduction

The North Anatolian Fault Zone (NAFZ) is one of the seismically most active strike–slip faults world-wide extending 1600 km from eastern Anatolia to the north Aegean Sea (Fig. 1a). The direction of slip corresponds well with Global Positioning System (GPS) derived 20 to 25 mm yr^− 1 westward motion of the Anatolian block with respect to Eurasia (McClusky et al., 2000, Reilinger et al., 2006). The Izmit earthquake of 17 August 1999 M_w = 7.4 exhibits a maximum surface displacement of 5.2 m at the Sapanca–Akyazi segment (Barka et al., 2002, Fig. 1b). Average coseismic slip obtained from teleseismic waveform inversion is 2.5 m (Tibi et al., 2001) and 2.9 m from strong motion records (Bouchon et al., 2002). Synthetic Aperture Radar interferometry (InSAR) data inversion (Wright et al., 2001) shows a maximum displacement of approximately 5 m near the mainshock and a total coseismic moment of 2.6 × 10²⁰ Nm. GPS data by Reilinger et al. (2000) indicate a geodetic coseismic moment of M₀ = 1.7 × 10²⁰ Nm and a maximum displacement at the Izmit segment of about 5.7 m. Delouis et al. (2002) identified four segments along the Izmit rupture by using combined GPS, SAR, teleseismic and strong motion data.

Lay and Kanamori (1980) studied body waves and surface waves of large earthquakes in the Solomon Islands region in an attempt to determine the stress distribution on the thrust plane. They found that relatively short-period seismic body waves are radiated from only small parts of the entire rupture plane, which generates longer-period surface waves and over which the aftershocks occur. They interpreted the results in terms of an asperity model. This asperity model is an outgrowth of laboratory experiments on rock friction. Byerlee (1970), Scholz and Engelder (1976) suggested that two sides of a fault are held together by areas of high strength, which they termed asperities. Extending this model to earthquake faults, Lay and Kanamori (1981) called the areas on the fault plane from where relatively short-period seismic body waves are radiated the fault asperities; it is assumed that the stronger spots are responsible for high-frequency seismic radiation.

From rock-deformation experiments in the laboratory three stages in acoustic emission fingerprints characterizing the failure of asperities are observed (Lei, 2003). During the first loading stage, the event rate which increases is documented by the inverse Omori law (Utsu, 1961). Simultaneously the b-value sharply decreases at the edge of the asperity. In the second stage prior to the mainshock, large magnitude events appear at the edge of the asperity (very close to the first stage hypocenters). During the third stage, significant increase and subsequent gradual decrease of b-value is observed along with a decreasing event rate (Zang et al., 1998).

In microscale and macroscale environments the b-value seems to rely on crack or rupture densities, which themselves result from the amount of applied stress or pressure, type of material, and rupture dynamics. Wiemer and Katsumata (1999) presented a “rupture mechanics approach” that relates b-values, coseismic slip and stress drop. Prior to an earthquake they suggest that areas subjected to high stress show low b-values. According to this approach high slip regions correspond to regions of high stress drop and high b-values of aftershocks. In this approach calculated slip distributions based on seismological and geodetic records of the mainshock are related to the aftershock seismicity parameters b-values. Sobiesiak et al. (2007) suggested that inhomogeneities in seismogenic fault areas can be mapped using b-value. These areas can be attributed to potential asperities. Wiemer and Wyss (1997) described that areas of low b-values with a size range of 5–15 km correlate with asperities in the Parkfield and Morgan Hill sections of the San Andreas Fault (SAF). This hypothesizes that low b-values indicate asperities on the fault prior to a large earthquake is supported by the study of Wyss et al. (2004) showing that b-values on locked patches of the SAF near Parkfield are systematically lower than b-values on creeping patches. This observation is a direct consequence of the inverse relationship of b-value to the applied shear stress. Amitrano (2003) and Schorlemmer et al. (2005) show that the high-stress environment of locked fault patches is more likely to support future large earthquake occurrence. Westerhaus et al. (2002) analyzed SABONET (Sapanca–BOlu NETwork) data before the Izmit earthquake. The lowest b-values (~ 0.8) were located at the fault bend which runs through the epicenter of the 1999 Izmit mainshock. At the bend, a localized stress concentration is expected from numerical models of seismicity along asperities. The site of lowest b-values had been considered to be the most likely place for a major earthquake, a conclusion that was confirmed by the Izmit earthquake, with epicenter located about 13 km from the anticipated site. Aktar et al. (2004) who analyzed the first 45 days of the Izmit aftershock sequence detected three zones of relatively high b-values, two of which coincide with asperities revealed by Bouchon et al. (2002). Aktar et al. (2004) mainly focused on the area between the Yalova and Karadere segments (see Fig. 1) and their analysis ended 42 days before the Düzce earthquake occurred. In this study we focus on the eastern part of the Izmit rupture (Sakarya, Karadere and Düzce regions). We discuss the Izmit aftershock sequence between October 18, 1999 and November 12, 1999 (i.e. the 25 days prior to the Düzce mainshock M_w = 7.2). We analyze spatial variations of 2414 aftershocks with errors < 5 km (horizontal and vertical). Variations in b-values along the rupture and with depth are compared with stress level change and postseismic slip distribution, and are discussed in the context with the preparation process of the Düzce earthquake.