Testing the effects of basic numerical implementations of water migration on models of subduction dynamics

Abstract. Subduction of oceanic lithosphere brings water into the Earth's upper mantle. Previous numerical studies have shown how slab dehydration and mantle hydration can impact the dynamics of a subduction system by allowing a more vigorous mantle flow and promoting localisation of deformation in the lithosphere and mantle. The depths at which dehydration reactions occur in the hydrated portions of the slab are well constrained in these models by thermodynamic calculations. However, computational models use different numerical schemes to simulate the migration of free water. We aim to show the influence of the numerical scheme of free water migration on the dynamics of the upper mantle and more specifically the mantle wedge. We investigate the following three simple migration schemes with a finite-element model: (1) element-wise vertical migration of free water, occurring independent of the flow of the solid phase; (2) an imposed vertical free water velocity; and (3) a Darcy velocity, where the free water velocity is a function of the pressure gradient caused by the difference in density between water and the surrounding rocks. In addition, the flow of the solid material field also moves the free water in the imposed vertical velocity and Darcy schemes. We first test the influence of the water migration scheme using a simple model that simulates the sinking of a cold, hydrated cylinder into a dry, warm mantle. We find that the free water migration scheme has only a limited impact on the water distribution after 1 Myr in these models. We next investigate slab dehydration and mantle hydration with a thermomechanical subduction model that includes brittle behaviour and viscous water-dependent creep flow laws. Our models demonstrate that the bound water distribution is not greatly influenced by the water migration scheme whereas the free water distribution is. We find that a bound water-dependent creep flow law results in a broader area of hydration in the mantle wedge, which feeds back to the dynamics of the system by the associated weakening. This finding underlines the importance of using dynamic time evolution models to investigate the effects of (de)hydration. We also show that hydrated material can be transported down to the base of the upper mantle at 670 km. Although (de)hydration processes influence subduction dynamics, we find that the exact numerical implementation of free water migration is not important in the basic schemes we investigated. A simple implementation of water migration could be sufficient for a first-order impression of the effects of water for studies that focus on large-scale features of subduction dynamics.