A study of ground motion attenuation in the Southern Great Basin, Nevada‐California, using several techniques for estimates of Qs, log A0, and coda Q

Abstract

As a first step in the assessment of the earthquake hazard in the southern Great Basin of Nevada‐California, this study evaluates the attenuation of peak vertical ground motions using a number of different regression models applied to unfiltered and band‐pass‐filtered ground motion data. These data are concentrated in the distance range 10–250 km. The regression models include parameters to account for geometric spreading, anelastic attenuation with a power law frequency dependence, source size, and station site effects. We find that the data are most consistent with an essentially frequency‐independent Q and a geometric spreading coefficient less than 1.0. Regressions are also performed on vertical component peak amplitudes reexpressed as pseudo‐Wood‐Anderson peak amplitude estimates (PWA), permitting comparison with earlier work that used Wood‐Anderson (WA) data from California. Both of these results show that Q values in this region are high relative to California, having values in the range 700–900 over the frequency band 1–10 Hz. Comparison of ML magnitudes from stations BRK and PAS for earthquakes in the southern Great Basin shows that these two stations report magnitudes with differences that are distance dependent. This bias suggests that the Richter log A0 curve appropriate to California is too steep for earthquakes occurring in southern Nevada, a result implicitly supporting our finding that Q values are higher than those in California. The PWA attenuation functions derived from our data also indicate that local magnitudes reported by California observatories for earthquakes in this region may be overestimated by as much as 0.8 magnitude units in some cases. Both of these results will have an effect on the assessment of the earthquake hazard in this region. The robustness of our regression technique to extract the correct geometric spreading coefficient n and anelastic attenuation Q is tested by applying the technique to simulated data computed with given n and Q values. Using a stochastic modeling technique, we generate suites of seismograms for the distance range 10–200 km and for both WA and short‐period vertical component seismometers. Regressions on the peak amplitudes from these records show that our regression model extracts values of n and Q approximately equal to the input values for either low‐Q California attenuation or high‐Q southern Nevada attenuation. Regressions on stochastically modeled WA and PWA amplitudes also provides a method of evaluating differences in magnitudes from WA and PWA amplitudes due to recording instrument response characteristics alone. These results indicate a difference between MLWA and MLPWA equal to 0.15 magnitude units, which we term the residual instrument correction. In contrast to the peak amplitude results, coda Q determinations using the single scatterer theory indicate that Qc values are dependent on source type and are proportional to ƒp, where p = 0.8 to 1.0. This result suggests that a difference exists between attenuation mechanisms for direct waves and backscattered waves in this region.