Joint Inversion of Lg Q Model and Site Response in the Middle East

We have developed a frequentist and a Bayesian method that uses two-station (TS) and reverse two-station (RTS) amplitude changes to construct an Lg attenuation and station site response model for the Middle East. TS and RTS have commonly been used to study Lg Q and relative site response. The TS Q and amplitude reduction estimates are a function of both the path-based attenuation (1/Q) as well as the site term (since TS do not account for differences in site response). The RTS-based Q and amplitude reduction estimates are only a function of the path-based attenuation while RTS relative site term estimates are only a function of the site response. Combining the estimates from the two methods can take advantage of the more reliable RTS measurements and the better spatial coverage of the TS measurements. Using more than 100,000 Lg waveforms of 2900 events recorded at 900 stations in the Middle East, we have developed a frequentist method and a Bayesian method to jointly invert for the Lg Q structure and relative site response from both TS and RTS measurements. Our frequentist method uses interstation path Q measurements and relative site response measurements through an LSQR algorithm while our Bayesian approach combine amplitude reduction measurements and relative site response measurements via a Bayesian framework. To evaluate both methods for the simultaneous inversion of TS and RTS measurements, we used Lg TS and RTS path geometries taken from our actual dataset to create simulation data from a checkboard Q model with the low value of 100 and the high value of 300, and station site terms with a random pattern ranging over two orders of magnitude. Both methods show excellent recovery of the true Q model for the Anatolian-Iranian Plateau. We will further apply both approaches to our actual data in the Middle East.