1. Introduction
Coda waves have been extensively used in seismology to study the scattering and attenuation properties of the lithosphere. The theoretical models proposed include single scattering (Aki and Chouet, 1975), diffusion (Aki and Chouet, 1975; Dainty and Toksoz, 1977), and energy flux (Frankel and Wennerberg, 1987). These models assume that the coda consists of scattered waves that have sampled and traversed a broad region surrounding the source and the receiver. The coda is thus thought to yield a spatial average of the Earth's attenuation properties over a large volume (Aki and Chouet, 1975). Spatial or temporal variations in coda decay rate could indicate a change in seismic properties, such as attenuation and the state of stress. Variations in coda decay rate have been suggested as earthquake precursors (Jin and Aki, 1986; Sato, 1988). It has been observed that b-value and Qc variations have the same trend around major earthquakes and an opposite trend around volcanic eruptions (b-value is a parameter that relates the magnitude with its frequency of occurrence).(Wyss, 1985; Novelo-Casanova et al., 1985; Sato, 1986;Jin and Aki, 1986). These observations have been explained
assuming that the b-value increases or decreases with the closing of crustal cracks according to the change of stress (Herraiz and Espinoza, 1986). If the increase of Qc (decrease of Q 21) is related to the closing of cracks, this parameter may be a more reliable indicator of tectonic stress than b(Aki, 1985). Temporal variations in the decay rate of coda amplitudes may measure the state of stress accumulation preceding a large earthquake (Chouet, 1979; Sato, 1988). Jin and Aki (1986) reported coda Q values about three times lower during the three-year period preceding the Tangshan earthquake of 1976 than afterwards. They also observed a similar temporal behavior in coda Q before and after the Haicheng earthquake of 1975.