GeoBIPy – Geophysical Bayesian Inference in Python

Sources/Usage: Public Domain. Visit Media to see details.

GeoBIPy – Geophysical Bayesian Inference in Python – is an open-source algorithm for quantifying uncertainty in airborne electromagnetic (AEM) data and associated geological interpretations. This package uses a Bayesian formulation and Markov chain Monte Carlo sampling methods to derive posterior distributions of subsurface electrical resistivity based on measured AEM data. Uncertainty estimates for the distribution of resistivity values can provide useful insights into layer boundaries and properties in the subsurface, and accurately conveys spatial variability in parameter uncertainty. After quantification of resistivity values, a posterior clustering approach is used to produce a geological classification of results into user-specified classes, where each class is represented by a distribution of possible resistivity values. At every point within the survey area, the probability of each class is determined, providing an estimate of geological model structural uncertainty. The current implementation is applied to time and frequency domain electromagnetic data. Future efforts will extend this algorithm to other types of geophysical data. 

To access the software: https://github.com/usgs/geobipy

References: 

Foks, N. L., and Minsley, B. J. (2020), GeoBIPy - Geophysical Bayesian Inference in Python. 10.5066/P9K3YH9O, available at https://github.com/usgs/geobipy. 

Minsley, B. J. (2011), A trans-dimensional Bayesian Markov chain Monte Carlo algorithm for model assessment using frequency-domain electromagnetic data. Geophys. J. Int. 187, 252–272. 10.1111/j.1365-246X.2011.05165.x 

Minsley, B. J., N. L. Foks, and P. A. Bedrosian (2020), Quantifying model structural uncertainty using airborne electromagnetic data, Geophysical Journal International, ggaa393, https://doi.org/10.1093/gji/ggaa393.