Based on this postulate, we propose a solution for generating self-consistently strain localization in geodynamical models: to describe the evolution of heterogeneity and anisotropy in mechanical behavior produced by sub-mesh scale processes. To do so, we propose to replace the traditional homogenization (averaging) approaches by a stochastic description of the mechanical behavior of rocks (Figure). This “revolution” is inspired by developments in meteorology and oceanography in the last two decades. Instead of searching for a deterministic rheology that produces the desired behavior (i.e., strain localization), we will explicitly model the evolution of spatial fluctuations in the rheology in response to changes in the microstructure (and composition) during deformation and analyze how small-scale heterogeneity translates into strain localization at larger scales. This formalism will account for feedbacks between deformation and other physico-chemical processes at imbricated spatial scales.
