Spectral amplitudes of regionally recorded Lg waves are studied in detail between 0.6 and 10 Hz, using vertical-component, velocity seismograms of the Eastern Canada Telemetered Network stations and a supplementary Seismic Research Observatory-type station at Glen Almond (GAC), Québec. We find that the site responses vary among these stations by more than a factor of 3 within the frequency range of interest. Furthermore, they are found to be strongly frequency dependent. Consequently, it is essential that they be taken into consideration in studies of Lg wave attenuation and Lg source spectra of regionally recorded seismic events.
We present a new method of measuring interstation surface wave attenuation, which is closely related to the conventional two-station method. While retaining all the desirable features of the conventional two-station method, the new technique, which we will call the “reversed two-station method,” allows simple, direct (one-parameter) determination of the Lg wave attenuation from sparse spectral data in a manner unaffected by station site effects and associated instrument error.
The reversed two-station method is successfully tested over weakly attenuating, short (53 to 210 km) interstation paths in eastern Canada, a difficult experimental condition by normal standards. The Lg attenuation coefficient (0.6 to 10 Hz) in eastern Canada is found to be frequency dependent and of the form γ(f) = 0.0008 f0.81 km−1. At higher frequencies, the Lg attenuation appears to be essentially frequency independent. This latter finding is preliminarily interpreted as evidence that regionally recorded Lg waves in the Canadian Shield are, as in the case of Lg waves propagating through the structurally complex Appalachian Province, contaminated by the high-frequency coda of Sn waves. The Lg contamination over the shield paths becomes severe starting at 14 Hz, twice the frequency above which the Lg signal propagating over the Appalachian Province becomes completely dominated by non-Lg arrivals.