ABSTRACT Characterizing the information in earthquake source spectra requires three measures: seismic moment M0, apparent stress σa, and stress parameter ΔσB. We estimate σa and ΔσB for 42 Ridgecrest, California, earthquakes (4.0≤Mw≤5.4), using three-component records within 50 km to minimize path effects. We analyze the data in both the time and frequency domains. We account for the depth dependence of source velocity and density and calibrate the results using observations at a rock site. Time-domain analysis for σa reveals significant apparent crustal attenuation (∝r−1.6, in which r is the centroid distance) and large site amplifications. In the frequency domain, we estimate near-surface impedance as a function of frequency at each station. We conduct a grid search with F-tests to constrain a frequency-dependent crustal Q model (Q(f)=q0fα) and site attenuation constant κ0 for each station, assuming a ω−2 model. The global best model has q0=60, α=0.675, with κ0 ranging from 0.01 to 0.05 s. σa and ΔσB were estimated using corrected observations. The σa values from both time- and frequency-domain analyses are in excellent agreement, ranging from 0.09 to 2.7 MPa with a geometric mean of 0.59 MPa. ΔσB ranges from 0.27 to 6.9 MPa with a geometric mean of 1.8 MPa. The ratio of ΔσB and σa (∼3.0) suggests the source spectrum in this magnitude range is close to a single-corner spectral model. We find both σa and ΔσB increase quickly with centroid depth that cannot be explained with depth-dependent crustal attenuation. Geometric mean values for σaF and ΔσB for earthquakes with centroid depths of ≥6 km are 0.92 and 2.91 MPa, respectively, approximately fourfold the values for earthquakes with centroid depths <6 km. Considering the significant impact to near-fault strong ground motion, the cause of this sharp transition deserves further investigation.
Read full abstract