The focusing and defocusing of teleseismic P waves predicted by a known three-dimensional structure beneath the Nevada Test Site is calculated by dynamic raytracing and superposition of Gaussian beams. The 20 to 100 km scale lengths of this model, having velocity fluctuations of several per cent, account for a factor of 3 fluctuation in the azimuthal pattern of P amplitudes. Since similar sized scale lengths and intensities of velocity fluctuation are commonly resolved in three-dimensional inversions of mantle structure in other regions, focusing and defocusing of teleseismic P waves is likely to be a ubiquitous feature of every test site. Hence, network averages of mb that weight azimuthal windows in which P energy is either focused or defocused will tend to either over- or underestimate the yields of underground tests. The results obtained with a known Nevada Test Site structure suggest that structural inversions having similar structural resolution may be capable of reducing 25 per cent of the variance in amplitudes due to the focusing and defocusing of upper mantle structure near the source.
Assuming that scattering processes are concentrated in the crust and upper-most mantle beneath the source and receiver, the integrated energy in the P coda should be more stable than direct P because it tends to homogenize the effect of focusing/defocusing structure near the source for scattering into direct P near the source. For direct P scattering near the receiver, however, even the late coda is not capable of completely removing the effects of focusing/defocusing near the source.