Description
We present a public domain, open-source, object-oriented python package called GeoBIPy, Geophysical Bayesian Inference in Python, a parallel enabled trans- dimensional Markov chain Monte Carlo inversion framework for geophysical data. GeoBIPy optimally computes a subsurface physical property model that fits the measured data, in addition to an ensemble of possible models, thus providing an estimation of uncertainty at different locations within the model. GeoBIPy uses the parallel-enabled python packages mpi4py and h5py in order to leverage large-scale distributed memory architectures. We show that large datasets taking over 3 months to invert can be completed in 2.5 days using 2016 cores.Presenter(s): Speaker: Leon Foks, Apogee Engineering contracted to United State Geological Survey Speaker: Burke Minsley, United States Geological Survey