THE EFFECT OF THE EARTHQUAKE RADIATION-PATTERN ON M(B) - A STUDY USING AFTERSHOCKS IN THE 1976 GAZLI SEQUENCE

We use published focal mechanisms to estimate radiation coefficients t o four short-period arrays recording teleseismic P from 38 aftershocks in the 1976 Gazli, Uzbekistan, earthquake sequence. We divide the obs erved P-wave amplitude by its radiation coefficient to estimate the P- wave amplitude that would be observed if it was from the maximum of th e double-couple radiation pattern. We use this new P-wave amplitude to calculate a P-wave magnitude, m(b)(C), that is independent of the P-r adiation pattern if the focal mechanisms are without error. Analysis o f variance shows that the random error in m(b)(C) is reduced relative to that in the original P-wave magnitudes m(b)(O) and that this reduct ion is statistically significant at the 11% level. Further, analysis o f variance demonstrates that the radiation coefficients calculated fro m the focal mechanisms contain error but that this error is probably n ot large enough to mask the detection of the radiation effect in m(b)( O). Published averages of the logarithm of P-radiation coefficients al low an assessment of the differences in network-averaged m(b) due to t he radiation pattern of point earthquake and explosion sources. Networ k-averaged m(b) from a vertical strike-slip earthquake can differ from an explosion of similar scaler moment by as much as 1.0 m.u. (magnitu de units). However, this difference can be as little as 0.2 m.u. if th e earthquake mechanism is 30 degrees dip slip. We argue that, if m(b) is required to be independent of the earthquake mechanism, the most ap propriate network average is (m) over bar(b)(C) - 0.48.