Bayesian Deconvolution of Seismic Array Data Using the Gibbs Sampler

The problem of monitoring for low magnitude nu clear explosions using seismic array data under a Comprehensive Test Ban Treaty CTBT requires a capability for distinguishing nuclear explosions from other seismic events Industrial mining explosions are one type of seismic event that needs to be ruled out when trying to detect nuclear tests We consider a Bayesian approach to the problem of detecting ripple red mining explosions Seismic array data are expressed as multidimensional time domain con volutions of an unknown pulse function representing the ripple red delay pattern with unknown signal path e ects sequences on each channel which are assumed to follow independent AR processes Using the Gibbs Sampler and a proposal of Cheng Chen and Li for blind deconvolution we develop an approach to estimating the delay parameters and the unknown signal path e ects sequence at each sensor Results for a ripple red mining explosion recorded at the Arctic Experimental Seismic Station ARCESS will be presented Finally the implica tions of our method for monitoring a nuclear test ban treaty are considered