Body wave amplitude patterns and upper mantle attenuation variations across North America

Abstract

Summary. A broad band amplitude study of P‐and SH‐waves from deep South American events recorded at WWSSN stations in the United States provides constraints on upper mantle variations in attenuation. The events were selected to minimize contamination due to source complexity, radiation pattern, and source structure. The prevalent feature in the short‐and long‐period bands is that the SH amplitudes show the same regional variations as the P amplitudes, with the regional variations of short periods being similar, but enhanced relative to the variations at long periods. The amplitudes show some systematic regional difference between the East Coast and Rocky Mountain provinces, and both short and long periods are enhanced at the Gulf Coast and midwestern stations, probably due to thick sedimentary bed receiver structure. Since detailed receiver structures for individual stations are poorly known, we used the working hypothesis that the regional amplitude patterns are caused by lateral variations in upper mantle attenuation. Both a constant Q and a frequency‐dependent attenuation operator were considered, with the relative amplitudes and waveforms in short‐and long‐period bands used to estimate the acceptable range of Δt* in the constant Q model, or the acceptable range of the parameter τm, which describes the high‐frequency roll‐off of the absorption band in the frequency‐dependent model. Time domain modelling indicates that the short‐period P and SH amplitudes and waveforms in the 1–10s period range do not unambiguously demand frequency‐dependent attenuation, allowing for the uncertainty in source frequency content and long‐period absorption band amplitude. The short‐period P amplitudes can be fit with a range in Δt*α= 0.5 s or a range in τm from 0.001 to 0.25 s, with either single parameter variation consistent with the general short‐period SH amplitude behaviour. A technique for determining the absolute value of t* and possible τm values appropriate for the 5–20 s band was applied to selected data. A value of t*= 0.8 s and small τm were found for travel paths from the deep events to North America, and lower t* or large τm values were found for paths to southern African stations. Long‐period amplitude variations cannot be explained by frequency‐dependent models with only a single high‐frequency roll‐off of the absorption band, but values of Δt*β determined from the long‐period SH amplitudes do exceedingly well in predicting the long‐period P amplitudes under the assumptions of constant Q and t*β/t*β= 4.0. The total range in long‐period Δt*β is about 3 to 4 s. Though receiver effects may cause some of this variation between stations, the data make it clear that such effects must produce amplitude behaviour similar to that caused by regional variations in attenuation. It is shown that the Δt*β values determined from the long periods do not predict the short‐period amplitudes in detail, but the predictions are not prohibitively outside the range of the short‐period amplitude variations. Spectral analysis of the long‐period SH‐waves was also performed, yielding a slightly greater range of Δt*β than the time domain results, but the resulting values of Δt*β do not predict the amplitude variations well, indicating the large effects of long‐period receiver functions on both the amplitude and spectral behaviour. Copyright © 1981, Wiley Blackwell. All rights reserved