ORIGINS OF AMPLITUDE VARIATIONS IN SEISMIC DOUBLETS - SOURCE OR ATTENUATION PROCESS

Accurate estimates of spectral ratios of two doublets (i.e. similar se ismic events) occurring along the San Andreas Fault, in the Hollister region have been computed using a cross-spectral method. Both events o f each doublet occurred within a 24 hr interval. Although both wavefor ms of a doublet are remarkably similar, computed spectral ratios exhib it strong, but steady variations with frequency in the first arrivals. The attenuation process can explain neither the sign nor the order of magnitude of these variations. On the contrary, they are closely rela ted to the source mechanism: we show that the slope computed from the linear and coherent part of the spectral ratio is an estimation of the variation in the apparent rupture duration between the two events. On e doublet exhibits an azimuthal distribution of the spectral-ratio var iations that is interpreted as the result of a change in the position of the hypocentres, relative to the rupture area. For this doublet, lo cation of hypocentres varies only from 10 to 30 per cent of the source length: this points out that small changes in rupture kinematics can lead to significant variations in the spectral ratio of earthquake dou blets, i.e. in the amplitude of each event. For the other doublet, the variations are related to a change in source length. We also have com puted spectral ratios all along the seismograms, using a moving window technique. When an azimuthal distribution of the spectral ratio is ob served in the first arrivals, we notice a decrease of the spectral rat io all along the seismograms. The doublet with strong, but non-azimuth al variations of the spectral ratio in the first arrivals, does not in dicate such a decrease of the spectral ratio along the seismogram. Thi s implies that the variations of the coda decay are related not only t o temporal or spatial changes of attenuation but also to source parame ters. These variations have the same order of magnitude as those provo ked by a temporal change in coda Q of 10 per cent. They can occult tim e changes of spectral ratio, for instance in case the attenuation in t he crust would vary before a strong earthquake.