Progress in Kappa Estimation at Yucca Mountain

We have been studying attenuation parameter kappa using recordings from the Southern Great Basin Digital Seismic Network in support of site and engineering characterization of the Yucca Mountain area in southern Nevada. Earthquake magnitudes range from 0 to 4.4, with 18 of M>=3.0. Kappa is estimated from the residual spectral slope after removing an omega-squared source spectrum. The kappa estimate primarily affects the median expected ground motions; uncertainty in the estimate can strongly increase probabilistic peak ground motion estimates when facility lifespans of 10⁴ years and greater are considered. We find that it is rare to resolve corner frequencies >6 Hz, even with high SNR recordings. With the available data, if site kappa is estimated solely from spectra with resolved error minima, the results can be biased toward lower kappa and stress drop estimates. To better constrain kappa, two limiting estimates are made from each spectrum, one assuming a corner frequency (cf) of zero, and the other, that it is infinite. For M>3 events, cf=0 corresponds to the original Anderson and Hough (1984) definition of kappa as the asymptotic high-frequency slope parameter. For M<1 events, cf=inf exploits the flatness of the displacement spectral slope well below the corner frequency. Any observed slope under these conditions is caused by attenuation and modeled by kappa. Finite corner frequency effects cause these estimates to under- and over-estimate kappa, respectively. Limiting estimates bound kappa to a ∼15 msec range, within which are found most of the resolved estimates from fitting. Two lines of evidence suggest that some variability in kappa must originate near the source. First, displacement slope kappa estimates from small earthquakes from a source region of 1-2 km square vary by over 20 msec. The source region is small enough that path or near-station variations are unlikely. Kappa variation may be due to near-source scattering or very low stress drops (<0.1 MPa). Second, inversions that include a source term in the kappa model consistently confirm its statistical significance. Estimates of site mean kappa depend on the method of distance correction. Over the distance range considered, the distance correction is weakly-constrained. A bilinear model that fits reasonably well indicates an average kappa for selected Yucca Mountain stations near 30 msec. Improved estimates and analysis considering energy and stress-drop constraints are being developed.