Open AccessIncorporation of probabilistic seismic phase labels into a Bayesian multiple‐event seismic locator
Author(s) -
Myers Stephen C.,
Johannesson Gardar,
Hanley William
Publication year - 2009
Publication title -
geophysical journal international
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.302
H-Index - 168
eISSN - 1365-246X
pISSN - 0956-540X
DOI - 10.1111/j.1365-246x.2008.04070.x
Subject(s) - event (particle physics) , prior probability , joint probability distribution , bayesian probability , probabilistic logic , computer science , a priori and a posteriori , algorithm , markov chain monte carlo , maximum a posteriori estimation , data set , set (abstract data type) , posterior probability , statistics , mathematics , artificial intelligence , physics , maximum likelihood , philosophy , epistemology , quantum mechanics , programming language
SUMMARY We add probabilistic phase labels to the multiple‐event joint probability function of Myers et al. that formerly included event locations, traveltime corrections and arrival‐time measurement precision. Prior information on any of the multiple‐event parameters may be used. The phase‐label model includes a null label that captures phases not belonging to the collection of phases under consideration. Using the Markov‐Chain Monte Carlo method, samples are drawn from the multiple‐event joint probability function to infer the posteriori distribution that is consistent with priors and the arrival‐time data set. Using this approach phase‐label error can be accessed and phase‐label error is propagated to all other multiple‐event parameters. We test the method using a ground‐truth data set of nuclear explosions at the Nevada Test Site. We find that posteriori phase labels agree with the meticulously analysed data set in more than 97 per cent of instances and the results are robust even when the input phase‐label information is discarded. Only when a large percentage of the arrival‐time data are corrupted does prior phase label information improve resolution of multiple‐event parameters. Simultaneous modelling of the entire multiple‐event system results in accurate posteriori probability regions for each multiple‐event parameter.