Bayesian inversion of MDJ seismograms for source parameters of contained DPRK underground nuclear explosions

Abstract

Abstract Using seismograms recorded at Mudanjiang (MDJ) seismic station in China, I propose a new approach to obtain source parameters of the five Democratic People’s Republic of Korea (DPRK) contained underground nuclear explosions. The contained events were located within about 1 km of each other at the DPRK Punggye-ri nuclear test site at Mount Mantap, a granite peak. The Green’s function from Punggye-ri to MDJ is assumed to be the same for each source. Each source is modelled as a spherical cavity in a homogeneous isotropic elastic full space, with four independent parameters constrained by published data on the properties of granite and analysis of the recorded MDJ seismograms. The effect of the ground surface is included as a planar reflection that modifies the pressure at the cavity boundary. The Green’s function for each event is estimated by deconvolving the seismogram for the estimated source time function. Very fast simulated annealing (VFSA) is used to search the parameter space to minimise the root-mean square difference among the estimated Green’s functions and their mean. The estimated Green’s functions are not identical, but are similar, and differ in amplitude by less than a factor of 4. The estimated source time functions may be used to obtain Green’s functions from Punggye-ri nuclear test site to other seismic stations by deconvolution of the corresponding seismograms. Mapping the nonlinear zones surrounding each explosion indicates that this part of the site was destroyed by the underground nuclear tests before its official destruction on 24 May 2018.