Seismic Source Functions and Attenuation from Local and Teleseismic Observations of the Nts Events Jorum and Handley

Several strong-motion seismograms recorded at 8 km from a large nuclear test at Pahute Mesa, Nevada Test Site, are modeled using the Cagniard-de Hoop technique. The ratio of vertical to radial motions suggest that the peak values are produced by ray paths that penetrated to a depth several kilometers below the source. A homogeneous layered Earth model with velocity increasing with depth was used in the modeling of the velocity time histories. The upper portion of the velocity model was determined by averaging bore-hole data and the lower portion was obtained from regional refraction measurements. Assuming a modified Haskell (1967) source representation, 1[,{ t) = 1/10 [1 e -K'(1 + Kt + (Kt) /2 8(KW)] we obtain a range of source descriptions with o/o varying with K and 8, o/o ( K, 8). The range of source models for Jorum are l{;o (5, 1) = 3.1, o/o (5, 2) = 1. 7, and l{;o (5, 3) = 1.2 times 10 cm , respectively. Given an explosion source description, it is a straightforward task to determine the teleseismic attenuation from WWSSN observations. From both the shortand long-period observations from these events, an average t * of 1.3 is obtained for compressional waves of a dominant 1-sec period. This estimate is insensitive to the range of K and B obtained from the near-field modeling. INTRODUCTION In recent years, a number of authors have compared sources derived from local strong motion data with teleseismic observations. In the case of explosions, investigators have examined the frequency content of short-period P waves to measure attenuation (e.g., Frasier and Filson, 1972). They estimate t, *to be about 0.5 where t, *=~with Tthe travel time of compressional (a) waves and Q, the quality factor. If t* is known along some ray path, then a convolution operator A (r, t*) can be constructed to correct a seismic pulse for attenuation [Carpenter et al. (1967)]. In the case of earthquake data, occasionally both longand short-period P and S waves at teleseismic distances and well-recorded localS waves are available. The long-period pulses are easily modeled synthetically. For example, the results of Burdick and Mellman (1976) for the Borrego Mountain earthquake indicate that the direct P wave actually contains P, pP, and sP, with the latter phase dominating. Modeling the phases sP and sS from the Borrego Mountain earthquake, Burdick (1978) estimated t11* to be 5.2 where the f3 refers to shear waves. Heaton and Heimberger (1977) modeled the strong-motion data and found that Burdick's teleseismic description of Borrego was compatible with the local observations. In general, comparing seismic pulses at various locations produced by earthquakes with the intent of determining Q is particularly difficult because of source finiteness and associated directivity effects. The complex radiation pattern associated with earthquakes introduces large uncertainties in comparing waveforms from various 51 52 DONALD V. HELMBERGER AND DAVID M. HADLEY stations. With the goal of avoiding this problem, we have reworked some of the best data available for comparing observations made near large nuclear explosions with teleseismic measurements of shortand long-period P waves. NEAR-FIELD STUDIES The megaton events, Handley and Jorum, considered in this study were located on the Pahute Mesa of the Nevada Test Site (NTS), see Figure 1. The near-field data described by Peppin (1974) were recorded at several azimuths at a distance of 8 km. These data provide a reversed profile with J orum shooting west and Handley toward the east. For this size event, recordings at 8 km probably better represent the source than would closer recordings. Near-source effects such as triggered movement on nearby joints should not dominate the observations and the recording site is less likely to be in the zone of nonlinear deformation. Seismograms recorded