DIFFERENTIAL ANALYSIS OF CODA-Q USING SIMILAR MICROEARTHQUAKES IN SEISMIC GAPS .1. TECHNIQUES AND APPLICATION TO SEISMOGRAMS RECORDED IN THE ANZA SEISMIC GAP

A large but conflicting literature exists that suggests that the scatt ering properties of the Earth in seismogenic areas may be temporally v ariable and that this variability could be correlated with the nucleat ion times of moderate to large earthquakes. To search for temporal var iations in coda Q in the vicinity of an expected moderate earthquake n ucleation region along the San Jacinto fault zone, pairs and clusters of earthquakes recorded by the Anza Seismic Network between 1982 and 1 992 are examined. To minimize differences arising from path and source variability, only the most similar examples of microearthquake wave f ields are intercompared. To assess coda Q differences between similar microearthquake signals, we search for systematic temporal changes in moving-window spectral amplitude ratios. For a single-scattering coda model with an envelope decay function of the form t(-m)e(-pi ft/Qc), t he relationship between the natural log amplitude spectral ratio and D elta Q(-1) = Q(c2)(-1) - Q(c1)(-1) is linear; i.e., ln r(f, t) = pi ft Delta Q(-1)(f) + W(f), where W(f) is a time-independent relative sour ce term. To obtain robust uncertainty estimates of Delta Q(-1)(f) and W(f), we utilize multi-taper spectral analysis coupled with a nonparam etric Monte Carlo confidence interval estimation procedure. For the mo st similar events in the Anza region (coda cross-correlation values of approximately 0.7 for 16 sec of coda), we find that coda Q as a funct ion of time is stable at a typical 1 sigma resolution of only between -25% and +50% of the reference coda Q value for the area, even in the best-constrained frequency bands. Comparison of similar microearthquak e coda signals with both short- and long-temporal separations indicate s that all of this variability can be reasonably attributed to random fluctuations in the coda that are driven by source variability and is not indicative of any systematic temporal variability in coda Q. To ob tain estimates of relative coda Q that are more precise than this leve l, the spectral ratio analysis technique requires significantly more s imilar earthquakes than are produced by seismogenic processes in the A nza region. Such earthquakes have been observed to occur in other regi ons of the San Andreas Fault system and are analyzed in Antolik et al. (1996).