Multisite estimation of explosive source parameters

Short-period P waves from five presumed explosions in eastern Kazakh with magnitudes mb ranging from 5.3 to 5.7 are examined at five teleseismic arrays: Lasa, YKA, OONW, WRA, and GBA. Time domain transfer functions that convert the lower-magnitude events to the largest event at the individual sensors of each array are calculated. These filters show that the observed source radiation of each event is space stationary over the aperture of each array but varies from array to array. At the Large-Aperture Seismic Array (Lasa) these transfer functions qualitatively match theoretical transfer functions based on Blake's solution for the seismic radiation from a cavity excited by a step function in pressure, in which the cavity radius for each event is scaled according to the Lasa magnitude. When all the arrays are considered simultaneously, apparent anisotropic radiation is seen in two of the five events, where an amplitude variation up to a factor of 3 is observed between two array sites. Multisite estimation of the source parameters is attempted in the frequency domain, in which Haskell's explosion spectral model is fitted to each event spectrum after corrections for instrument response and attenuation are applied. At each array the attenuation that allows the fitted parameters to vary as dictated by the model is chosen as correct. With attenuation estimated at each array, the spectra observed at all arrays for a single event are simultaneously fitted to the three parameters of Haskell's source model. For most arrays the individual and simultaneous fitting schemes yield reasonable values for the source parameters. The assumed source model and the estimated attenuation parameter for central Asia to Lasa apparently explain why the spectral ratio of high- to low-frequency energy in explosion sources increases as yield decreases.