Uncertainty analysis of potential field data inversion

Figure Caption

A set of recovered density models range from a wide spectrum of model features with the depth slices at -2000 m (upper) and cross sections at 4790800 m (lower). We note that these models contain the consistent level of data fitting. The diverse model characteristics reveal the uncertainties of gravity gradient inversions.

Research Summary

We have developed an empirical method to quantify the uncertainties of the inverted models in the 3D deterministic inversion framework. We have employed the mixed Lp norm inversion, a very recently developed inversion method, which allows different norms to be applied to the four different components of the regularization term. Our work shows that the uncertainties of the 3D inverted models can be empirically assessed in the deterministic framework by taking advantage of the adjustable parameters. We emphasize that a better practice for interpreting field data sets is to create a large sequence of equivalent yet diverse models and factor the uncertainties into interpretations, instead of basing his or her interpretation on a single “optimal” model.