Regional Variations of the Cutoff Depth of Seismicity in the Crust and Their Relation to Heat Flow and Large Inland-Earthquakes
Total Page:16
File Type:pdf, Size:1020Kb
J. Phys. Earth, 38, 223-250, 1990 Regional Variations of the Cutoff Depth of Seismicity in the Crust and Their Relation to Heat Flow and Large Inland-Earthquakes Kiyoshi Ito * Regional Center for Earthquake Prediction, Faculty of Science, Kyoto University, Takatsuki 569, Japan More than 8,000 earthquakes have been relocated to derive regional variations of the seismic-aseismic boundary in the mid-crust of the northern Kinki district of Japan. The boundary depths are estimated as 13-15 and 18-20 km in the southwestern and northeastern parts of the study area, respectively. The relationship between the seismic cutoff depth and the cause of large intraplate earthquakes is studied, making use of the present observations and other available data of seismicity and surface heat-flow, on the basis of the brittle-ductile transition of rock deformation. The regional variations in the cutoff depth of seismicity appear to be well correlated with the thermal structure of the crust. The cutoff depths in various heat-flow provinces in Japan and other countries are found to be inversely proportional to the surface heat-flow values, with the depths roughly corresponding to isotherms of 200-400•Ž. The shape of the depth-frequency distribution of earthquakes calculated from high-quality data is quite similar to that of the shear resistance calculated using a simple brittle-ductile transition model. Large intraplate earthquakes appear to originate at the peak or just below the peak in the depth-frequency distribution, which also corresponds to the deepest portion of the seismogenic layer. Furthermore, in the source area of large earthquakes, rupture seems to start where sharp changes occur in the cutoff depth of seismicity. Thus, the seismic-aseismic boundary is closely related to large intraplate earthquakes and, in turn, to the tectonics of island arcs. 1. Introduction Seismic activity in the crust is mostly confined to the upper crust with only a few earthquakes occurring in the lower crust. This fact was first noted during the late 1960's in California (i.e., Eaton et al., 1970). In Japan, the restriction of seismicity to the upper crust was first reported for the eastern Chugoku and northern Kinki districts (Hashizume, 1970a ,b). Since then, it has been recognized in many other regions through the Received March 16 , 1990; Accepted July 14, 1990. * Present address: Research Center for Earthquake Prediction , Disaster Prevention Research Institute, Kyoto University, Uji 611, Japan. 223 224 K. Ito construction and improvement in microearthquake observation networks under the Earthquake Prediction Project (Oike, 1975; Takagi and Hasegawa, 1977; Takagi and Matsuzawa, 1988). To date, many papers have been published on seismic activity with reference to the distribution of active faults in Japan, but extensive studies of the seismic depth distribution are rare. Kobayashi (1977) first interpreted the cutoff depth distribution of seismicity as the brittle-ductile boundary of rock deformation, indicating that surface heat-flow values are related to the seismic depth in southwestern Japan. This idea was mainly based on • the results of laboratory experiments of rock deformation. He also suggested the possibility that the crust was not made of two layers, (i.e., 'granitic' and 'basaltic' layers), but was composed of a rock such as granodiorite. However, the seismicity data used in his paper were not very accurate, and further studies did not develop for a long period of time. On the other hand, the seismic-aseismic boundary was found to be clearly defined along the San Andreas fault (Eaton et al., 1970) and the causes of the boundary have been studied in relation to the results of laboratory experiments of rock deformation (Byerlee, 1968; Brace and Byerlee, 1970). On the basis of the accumulated seismic data and rock experiments, a conceptional rheologic model of an earthquake occurrence in the crust has been presented by Brace and Kohlstedt (1980) and Sibson (1982). The model has been applied to the seismic depth distribution data from several regions in California, Europe, and other countries (Meissner and Strehlau, 1982; Chen and Molnar, 1983; Doser and Kanamori, 1986; Mikumo et al., 1989). According to Sibson (1984), the cutoff depth of seismic activity varies along the San Andreas fault, and large shocks are likely to occur at the bottom of the seismogenic layer. Moreover, the tectonic features of the continents and island arcs have been discussed in terms of the strength of the lithosphere derived from the thickness of a brittle layer (Vink et al., 1984; Shimamoto, 1989). Therefore, the precise determination of hypocenters can be considered as very useful in revealing their relation to the occurrence of large intraplate earthquakes and consequently to tectonic features. In this paper, hypocenters of a large number of earthquakes are relocated in order to examine regional variations of the seismic-aseismic boundary in the northern Kinki district, where a dense network has been monitoring seismic activity. The relationships of the boundary to the thermal structure and large intraplate earthquakes are also examined, along with the relationship between the depth of the seismic-aseismic boundary and surface heat-flow data for well-determined hypocenters found in Japan and other countries. From these data, the occurrence of crustal earthquakes is discussed on the basis of the rheologic model of rock deformation. Other evidence is presented which supports the model results, such as variations of the seismic-aseismic boundary near active volcanoes where abrupt changes occur in the thermal structure. Further, the depth of the seismic-aseismic boundary is shown to be closely related to occurrences of large earthquakes or the tectonics in the inner zone of the Japanese Islands. 2. Earthquake Data and Hypocenter Determination in the Northern Kinki District Seismic activity in the northern Kinki district has been monitored with a dense J. Phys. Earth Regional Variations of the Cutoff Depth o f Seismicity 225 Fig.1. Map showing the seismograph stations of the Abuyama Seismological Observatory, Kyoto University, in the northern Kinki district of Japan. network since 1963 by the Abuyama Seismological Observatory, Kyoto University (Okano and Hirano, 1968). In 1975, the network was improved and equipped with a telemetered system using telephone lines and radio links (Kuroiso and Watanabe, 1977). The system employs automatic data processing for phase detection and the determination of hypocenters and magnitudes (Watanabe and Kuroiso, 1977). Since then, both automatically and manually processed data have been stored over a period of more than ten years. The network stations which supply data used in this study are shown in Fig. 1. Although the automatically processed data are accurate enough for the real-time monitoring of seismic activity, high-quality data are necessary for the precise determination of focal depths. More than 21,000 earthquakes were located by manually examining the data and their phase data were stored during 1976-1987. For the present study, all earthquakes were relocated that satisfy conditions to be described later, making use of the manually picked P and S wave data. The P wave data with quality A and B were used for the relocation of hypocenters. The quality is determined from the uncertainty of P wave arrival time, within 0.02 s as A and 0.05 s as B. S wave arrivals are used only for estimating the starting value of origin time. Hypocenters and origin times were reset from P wave arrival times in order to minimize the standard error of P wave residuals between the observed and calculated travel times for a horizontally layered velocity structure. The model crustal structure used consists of three layers having P wave velocities of 5.5, 6.0, and 6.7 km/s with the Vol. 38, No. 3, 1990 226 K. Ito thicknesses of 5.0, 20.0, and 10.0 km, respectively. The model is based on the results of refraction surveys (Yoshii et al., 1974; Ito and Murakami, 1979). Since the study covers a region approximately 100•~100 km with the focal depths of earthquakes ranging over 5-20 km, most of the rays emitted from the hypocenters to the stations are estimated as direct waves. Therefore, the upper part of the crustal structure is mainly used for the calculation of travel times. Corrections for station heights are made when calculating the travel times by use of a P wave velocity of 4.5 km/s for the surface layer. The uncertainty in the determination of hypocenters from the present network data has been estimated by Ito and Kuroiso (1988). The effects of station corrections and the thickness of the surface layer have considerable influence in the estimation of the focal depths. Absolute focal depths have an error of about 2 km, due to the difference of 2 km in the thickness of the surface layer. According to Ito and Kuroiso (1988), however, the uncertainty in the relative focal depths calculated from this network data is one-tenth of the absolute depths regardless of the velocity structure; this uncertainty was determined by comparing the hypocenters to those determined using the master event location technique. In order to confirm the regional variations of focal depths, simultaneous determinations of the hypocenters and the three-dimensional velocity structure are also conducted using an inversion method. As a result, the hypocenters are accurate enough to derive regional variations of focal depths over the entire study area. The earthquakes selected for this study satisfy the following criteria: the P wave arrivals for a given earthquake must have been recorded at more than seven stations with the standard error of P wave residuals being within 0.1 s. It is important for precise focal-depth determination that the epicentral distance to the nearest station is comparable to focal depth and that the range of the epicentral distance is sufficiently large.