Importance of global spherical geometry for studying slab dynamics and evolution in models with data assimilation

Numerical modeling of subduction has proved increasingly important in understanding the evolution of Earth's internal dynamics and surface tectonic responses. A key criterion in judging models is their capability in matching both past plate tectonic constraints and the present-day mantle structures. In the past decades, most such models covered a regional mantle domain due to computational costs, while only recently sophisticated global spherical simulations became possible. Here we evaluate the effects of chosen model domain and applied boundary conditions in regional simulations on their resulting model results, by comparing them with more natural global simulations. All these models are based on the same sequential data-assimilation method and adopt the same plate tectonic history as surface boundary conditions. The only difference is that the regional models cover different geographic regions, representing a subset of the global one that covers the entire Earth. Besides illustrating the general impacts of chosen model setup on the resulting mantle evolution, we also make suggestions on the applicability of using regional models to replicate past subduction in four major circum-Pacific regions. The results show that the slabs from the global model match seismic tomography much better than the regional results in East Asia and South America; the main reason is that the regional models prohibit large-scale mantle flow as that in the global case, due to the non-penetrating side walls cutting off far-field forces. This limitation of regional models cannot be resolved by varying the model domain. In contrast, for North America and Fiji-Tonga, global and regional models produce similar slab structures, both matching tomography. This reflects that their fast-retreating overriding plates play an important role on the regional mantle flow and slab dynamics.