Isotropic and deviatoric characterization of the Coalora nuclear explosion in Yucca Flats

SUMMARY Near‐source data from the nuclear explosion Coalora detonated at Yucca Flats, Nevada Test Site (NTS), are utilized to constrain the seismic‐source function. The equivalent seismic source is interpreted in terms of physical processes in the source region with the aid of data from within the explosion's non‐linear region. The isotropic, deviatoric and spall contributions are separated and quantified. Standard spectral interpretations of the radiated wavefield for source parameterization are contrasted with complete waveform modelling with moment tensor determination. Individual waveform spectra (source‐receiver offsets < 2 km) can be interpreted in terms of an isotropic source model, which is in agreement with a Mueller‐Murphy model, including f −2 high‐frequency decay and source corner frequency of 1.8 Hz. The deviatoric component of the moment tensor is a factor of 5–10 times smaller than the isotropic component. Deviatoric source radius, as estimated from the spectral data, is 125 m, smaller than the equivalent elastic source radius, which is bounded between 133 and 202 m. Stress drop estimated with the Brune source model is 45 bars with an average slip of 17 cm. Moment tensor inversion produces an isotropic source strength of 8 × 10 20 dyne cm, while scalar moments from the spectral interpretation are a factor of 2.5 larger. This difference is attributed to the application of whole‐space propagation path corrections with a free‐surface amplification to the spectral interpretation. The spall source is longer period and delayed in time from the initial explosion. Its contribution to the diagonal elements of the moment tensor is dominant on the M zz component, a factor of 3 larger than the M yy and M xx components. Spall source strength from waveform inversion is within a factor of 2 of forward models developed from acceleration data within the spall zone. It is longer in duration than the forward prediction, reflecting the effect of a quasi‐point source assumption in the forward model. Complex propagation effects extend in time and homogenize the data beyond 2 km as exemplified by wave trains at 5 km that are 20 s in duration and similarity of radial, vertical and transverse acceleration spectra. In contrast, observations at 2 km or less are short in duration with strong differences between transverse and radial‐vertical spectra. These apparent propagation path effects suggest that source biases can develop at ranges as close as 2–5 km.