Crustal Structures Inferred from Rayleigh-wave Signatures of Nts Explosions by Thomas

An improved method for determining plane-layered earth models that accurately represent the important features controlling the amplitude and wave form of surface waves is presented. The method includes a formal inversion of phase and group velocity data determined from observed seismograms and is applied to the Rayleigh waves from Nevada Test Site (NTS) explosions recorded at Albuquerque, New Mexico and Tucson, Arizona. For both paths the observed dispersion agrees with that from the models with a maximum residual of only O.01 km/sec. Further, the models are consistent with other available information about these paths (e.g., from refraction surveys). To properly account for local differences in the material at the source, an approximate theory is constructed in which the amplitude excitation is computed in a source structure and the dispersion in a separate path structure. Using this theory and the crustal models from the inversion, synthetic seismograms are computed that match the observed seismograms remarkably well. INTRODUCTION We present an improved method for determining plane-layered earth models that accurately represent the important features controlling the amplitude and wave form of the Rayleigh waves propagated along particular paths. Since dispersion data provide valuable information about earth structure, it is desirable to develop effective inversion techniques for interpreting them. Further, the better our models account for path effects, the more confidently we can relate the amplitudes of Rayleigh waves to source parameters, source structure, and dissipation effects. This is particularly important in problems related to the monitoring of underground nuclear explosions. As an application, we use Rayleigh wave recordings of Nevada Test Site (NTS) explosions at the WWSSN stations ALQ {Albuquerque, New Mexico) and TUC {Tucson, Arizona) to infer the crustal structure for the NTS-ALQ and NTS-TUC paths. These paths, or portions of them, have been previously studied using body waves (e.g., Prodehl, 1970; Warren, 1969; Langston and Helmberger, 1974) and surface waves {e.g., Keller et al., 1976; Alexander, 1963; Wickens and Pec, 1968) and these previous solutions provide a useful check, Our inversion method includes a direct determination of phase and group velocities from the recorded Rayleigh waves and a formal linear inversion of these data for earth structure. It is similar to previously applied techniques (e.g., Keller et al., 1976), but represents an extension and improvement. Since both phaseand group-velocity data are used, our data have considerably greater resolving power than do group velocities alone. Also, the formulation and solution of the inversion problem is done by a more efficient and flexible method than previously used. Further, an interesting corroboration of the models is that theoretical seismograms computed with them show remarkable agreement with the obServations. Both phase and group velocities can be rather easily determined from explosion recordings because the phase and group delays at the source are nearly zero. The 1399 1400 THOMAS C. BACHE, WILLIAM L. RODI, AND DAVID G. HARKRIDER methods used for determining group velocities (narrow band filtering) and phase velocities (unwrapping the phase spectrum) produce nearly independent estimates for the two. Nonetheless, we find the values for these paths to be quite consistent from event to event and in excellent agreement with the differential relationship between phase and group velocity. The earth models found by the inversion are rather simple and are consistent with refraction data where it is available. The NTS-TUC path is chm'acterized by a crustal thickness of 31 km and the average crustal thickness for NTS-ALQ is 42 km. The phaseand group-velocity data are all fit within 0.01 km/sec. The phasevelocity data are especially important if theoretical seismograms are to match the observations, as they do here. For example, a seismogram, its negative and its Hilbert transform all have the same group-velocity dispersion. To compute synthetic seismograms, we must address the fact that conventional surface-wave theories (e.g., Harkrider, 1964) cannot be used in a consistent way when events in close proximity occur in different source materials, as is common at NTS. Therefore, we begin by constructing, albeit in a somewhat a d h o c way, a theory in which two structures are used to model the source-receiver travel path. The amplitude excitation is computed in a source structure and the dispersion is computed in a separate path structure. A transmission coefficient accounts for passage of Rayleigh waves between the two. RAYLEIGH WAVES FROM PROXIMATE EVENTS IN DISSIMILAR SOURCE MATERIALS A computationally convenient formulation of the theory for the surface waves generated by a point source in a plane-layered earth model was given by Harkrider (1964, 1970). The formulation of the theory is entirely in terms of linear elasticity, although the effect of anelastic attenuation can be included via an empirically determined Q operator. The source representation may be in terms of elementary point forces (Harkrider, 1964) or a general expansion of the outgoing elastic waves in terms of spherical harmonics (Harkrider and Archambeau, 1978). For spherically symmetric explosions a convenient source representation is the reduced displacement potential ~(r) defined by