2017 Volume 123 Issue 9 Pages 733-745
Metamorphism refers to the reactions that proceed in rocks in response to the dynamic environmental factors of the Earth's interior, which include temperature, pressure, and chemical composition. Because it is difficult to obtain time-series data of metamorphic processes directly, it is necessary to extract information from the spatial patterns of the final states of the rocks (e.g., the textures of metamorphic rocks) to understand the conditions and processes of metamorphism. The complications of metamorphic rock textures range widely, from simple problems that can be modeled based on rigid theoretical backgrounds, to complex problems in which several processes interact in nonlinear ways, and forward models themselves are still being developed based on combinations of empirical laws. In this paper, we review recent progress in the analysis and modeling of metamorphic rock textures, with a particular focus on the thermodynamic analysis of zoned minerals and forward modeling of reaction-induced fracturing, employing the distinct element method (DEM). We show that stochastic inversion analyses based on Bayesian inference can be a powerful tool for solving various petrological problems characterized by parameters with undefined values and noise. By effectively using algorithms of machine-learning, the approach of data assimilation, which combines numerical simulations and observed data, is likely to yield a breakthrough in terms of deciphering the complex textures of metamorphic rocks.
