Four-dimensional multiscale global subduction models with data assimilation and realistic rheology using ASPECT

The solid Earth is a complex system characterized by dynamic interactions among various tectonic components. Global mantle convection models equipped with data assimilation can effectively reproduce past subduction and associated mantle flow, providing a realistic framework for evaluating the intricate dynamic processes within the solid Earth. Despite recent advancements of data assimilation methods, their widespread application has been hindered by high computational costs due to the need for increasing model resolution and nonlinear rheology. Here, leveraging the mantle convection code ASPECT, we develop a multiscale global mantle convection model featuring adaptive data assimilation and employing nonlinear visco-plastic rheology. Our model successfully reproduces complex mantle evolution and structures, consistent with both observational constraints and previous model results. This represents the first published global mantle flow model built using ASPECT to achieve Earth-like subduction, with the aid of nonlinear rheology and adaptive data assimilation. Furthermore, the incorporation of adaptive mesh refinement and high-order finite element ensures high resolution and accuracy of model results. These advancements will contribute to a better understanding of plate tectonics and continental evolution.