Digital ground-motion data recorded by the Western Canada Telemetered Network (WCTN) are used to examine the attenuation and source parameters of earthquakes in the Cascadia region of southwestern British Columbia and northwestern Washington. The data base is comprised of over 1000 vertical-component Fourier spectra, from earthquakes of magnitude 3 to 7 at distances from 10 to 500 km. Regression analyses determine the shape and coefficients of the regional attenuation curve and source spectra for 68 earthquakes. Seismic moment and stress drop are inferred from the amplitudes of the source spectra.
Shallow (h < 10 km) earthquakes have an attenuation curve with a complex shape, exhibiting significant flattening (no apparent geometric spreading) in the distance range from 75 to 230 km; this shape is probably a result of strong reflected phases from mid-crustal discontinuities and the subducting slab. Events within the subducting slab and the lower crust exhibit a simple R−1 attenuation curve. The anelastic attenuation coefficient for the region as a whole is given by Q = 380f0.39.
The duration of motion for each WCTN record is determined as the value that yields the observed relationship between time-domain and spectral-domain amplitudes, according to random process theory. These durations are approximately constant within 50 km of the source, then increase with distance as 0.07R.
The high-frequency ground motions from Cascadia earthquakes are relatively weak. Cascadia source spectra are characterized by an average Brune stress drop of about 30 bars. This is significantly lower than the average California stress drop of 70 to 100 bars, and dramatically lower than the average eastern stress drop of 150 to 200 bars. It is concluded that there are significant regional variations in source parameters. Hazard estimates for the Cascadia region based on California groundmotion relations may be overly conservative.