Comprehensive Magnitude Yield Estimation for Nuclear Explosions: A Maximum Likelihood General Linear Model (MLE-GLM88)

摘要

This study performs maximum likelihood event magnitudes and station effects for 111 nuclear explosions from more than 8 different test sites using a general linear model. The estimates are performed for the three P-wave phases a, b, and max by considering the signals, non-detection, and clipping information. The bias between a model that does not include censoring effects from non-detection and clipping is demonstrated. The maximum-likelihood magnitude estimates are performed using data for each phase individually and the three phases combined. The difference in mb's between the phases is studied for some of the major test sites. A presumed multiple event is found to be characterized with large max phase amplitude compared to the a phase, whereas in the case of a presumed cratering event showing surficial collapse, the amplitudes for the b and max phases are smaller instead. In addition, the biases are examined at several test sites that may exist for general linear models based on a specific test site relative to a model that contains sources from many test sites. The effects of assuming correlation among the events on the magnitude estimates using the present event-station distribution are investigated using an inter-event correlation model. Keywords: General linear model, Maximum likelihood, Yield estimation, Body waves. (JHD)