The Se ' Paration of Source , Site , and Path Effects on High Frequency Seismic Waves : an Analysis Using Coda Wave Techniques

This work represents from local earthquakes in study are to evaluate the further our understanding knowledge concerning the an intensive study of coda waves California. The goals of this ability of coda wave analysis to of the earth, and to augment our generation of the coda itself. The methods used here are the simple extensions of single station coda wave techniques used by many workers (e.g. Aki and Chouet, 1975). Under non-restrictive assumptions of coda stability, a method is constructed to isolate source or site effects on high frequency waves (1.5-24 Hz). These techniques allow incorporation of data from earthquakes of many sizes, and from stations lying great distances apart. Digital data are drawn from the U.S.G.S. California Network (CALNET) archives; over 1200 records are used from over 90 earthquakes in the Coast Ranges of California between San Francisco and San Luis Obispo. Application of our techniques yields a variance reduction of 75-90% depending on data processing details. The methods of Aki (1980a) and Aki and Chouet (1975) are also used in order to evaluate the path effect (Q) of both coda and direct shear waves. The site effect on coda waves is calculated for the data set which was especially collected for this purpose. In spite of a bias toward hard rock sites in the CALNET, a tremendous variation in site response is found. At low frequencies the results spread over a range that is too large to be explained solely by the impedence effect. Other processes such as inefficiently damped trapped modes must exist in extreme cases. The model of coda waves as backscattered waves from randomly-situated heterogeneities in the earth must be extended to include this possibility. At high frequencies attenuation controls the site effect, although strange variations, especially at granite sites, remain unexplained. The source effect is calculated for the data set in much the same manner as for the site study. Results are expected to be sensitive to details of the rupture process and anelastic properties of the near source medium. Variations attributable to near source attenuation exist, but are not as strong as the attenuation induced site response variations seen in Chapter 3. The results are compared to simple w-squared and w-cubed source models. The w-squared model fits the best, except for the group of Coyote Lake earthquakes that exhibit a constant corner frequency independent of size. This indicates that a source-controlled limiting corner frequency, or fmax, exists in the Coyote Lake data. The result is not strong and should be investigated further with a more appropriate distribution of data. An exhaustive search is carried out to determine the factors upon which the quality factor of coda waves (Qc) depends. The most interesting result is a dramatic change in crustal Qc (by a factor of 2) at high frequencies between the high Q Salinian and the low Q Franciscan regions. This could be due to scattering loss at high frequencies in the highly-deformed Franciscan, or to low intrinsic Q. Except for this crustal effect Qc is shown to be independent of source-reciever path. The quality factor of shear waves (Qp) is found to agree with crustal Qc within error bars. Temporal variations in Qc and ML-Moda are also investigated. ML-Mcoda is observed to decrease after the magnitude 5.0 Bear Valley Earthquake. I have demonstrated that coda analysis can effectively deal with an extensive data set. The results of this thesis encourage further work on many fronts. Thesis Supervisor: K. Aki Title: Professor of Geophysics