Effects of Source RDP Models and Near-source Propagation: Implication for Seismic Yield Estimation

Abstract

Ð It has proven dicult to uniquely untangle the source and propagation eects on the observed seismic data from underground nuclear explosions, even when large quantities of near-source, broadband data are available for analysis. This leads to uncertainties in our ability to quantify the nuclear seismic source function and, consequently the accuracy of seismic yield estimates for underground explosions. Extensive deterministic modeling analyses of the seismic data recorded from underground explosions at a variety of test sites have been conducted over the years and the results of these studies suggest that variations in the seismic source characteristics between test sites may be contributing to the observed dierences in the magnitude/yield relations applicable at those sites. This contributes to our uncertainty in the determination of seismic yield estimates for explosions at previously uncalibrated test sites. In this paper we review issues involving the relationship of Nevada Test Site (NTS) source scaling laws to those at other sites. The Joint Veri®cation Experiment (JVE) indicates that a magnitude (m b ) bias (dm b ) exists between the Semipalatinsk test site (STS) in the former Soviet Union (FSU) and the Nevada test site (NTS) in the United States. Generally this dm b is attributed to dierential attenuation in the upper-mantle beneath the two test sites. This assumption results in rather large estimates of yield for large m b tunnel shots at Novaya Zemlya. A re-examination of the US testing experiments suggests that this dm b bias can partly be explained by anomalous NTS (Pahute) source characteristics. This interpretation is based on the modeling of US events at a number of test sites. Using a modi®ed Haskell source description, we investigated the in¯uence of the source Reduced Displacement Potential (RDP) parameters w ¥ , K and B by ®tting short-and long-period data simultaneously, including the near-®eld body and surface waves. In general, estimates of B and K are based on the initial P-wave pulse, which various numerical analyses show to be least aected by variations in near-source path eects. The corner-frequency parameter K is 20% lower at NTS (Pahute) than at other sites, implying larger eective source radii. The overshoot parameter B appears to be low at NTS (although variable) relative to other sites and is probably due to variations in source conditions. For a low B, the near-®eld data require a higher value of w ¥ to match the long-period M S and short-period m b observations. This¯exibility in modeling proves useful in comparing released FSU yields against predictions based on m b and M S .