Abstract by Martha Morrise
Jared Whitehead, Ron Harris
Reconstruction of Historical Earthquake Events Using Markov Chain Monte Carlo Methods
Markov chain Monte Carlo (MCMC) methods are a class of algorithms for sampling from unusually shaped probability distributions. Using Bayes’ Theorem, these algorithms are also useful for solving inverse problems, which are situations in science that use the results of a system to infer the conditions that caused them. We apply an MCMC algorithm to historical tsunami observations from Indonesia to construct probability distributions for the location and parameters of earthquakes that caused them centuries ago, for which only anecdotal evidence exists. Our methods allow for an efficient and mathematically rigorous inference of earthquake source parameters, which in turn will allow for more accurate estimates of the current stress level on fault zones. Though incomplete, our preliminary results have been promising and have produced parameter estimations that improve upon previously accepted solutions of this inverse problem.