Falsification of magmatic intrusion models using outcrops, drillholes, and geophysics

Magmatic intrusions can host a variety of minerals and often serve as key targets in mineral exploration endeavors. During exploration, the manual construction of an “optimal” intrusion model based on the available data is a standard practice, which then forms the foundation for subsequent interpretations and decision-making processes. However, this optimal-model-to-interpretation paradigm that relies heavily on expertise presents a significant challenge in assessing the uncertainty of the inferred model. To address the problem, we have developed an innovative and automatic framework with a core component: falsification. The first step involves the stochastic construction of prior magmatic intrusion models using outcrop and drillhole data. We can also derive a prior distribution of the physical property (e.g., density contrast) for the intrusive body from rock sample measurements or textbooks. The second step attempts to falsify the prior geometric representations and the prior physical property, obtained from the first step, using geophysical data. The unfalsified prior models can serve as realistic and valid priori for diverse Bayesian inversion methods, resulting in more reliable quantification of the posterior uncertainty. Our study highlights the necessity and significance of falsification in the prior model construction.