John F. Rudge

Olivine‐hosted melt inclusions are an important archive of pre‐eruptive processes such as magma storage, mixing and subsequent ascent through the crust. However, this record can be modified by post‐entrapment diffusion of H + through the olivine lattice. Existing studies often use spherical or 1D models to track melt inclusion dehydration that fail to account for complexities in geometry, diffusive anisotropy and sectioning effects. Here we develop a finite element 3D multiphase diffusion model for the dehydration of olivine‐hosted melt inclusions that includes natural crystal geometries and multiple melt inclusions. We use our 3D model to test the reliability of simplified analytical and numerical models (1D and 2D) using magma ascent conditions from the 1977 eruption of Seguam volcano, Alaska. We find that 1D models underestimate melt inclusion water loss, typically by ∼30%, and thus underestimate magma decompression rates, by up to a factor of 5, when compared to the 3D models. An anisotropic analytical solution that we present performs well and recovers decompression rates within a factor of 2, in the situations in which it is valid. 3D models that include multiple melt inclusions show that inclusions can shield each other and reduce the amount of water loss upon ascent. This shielding effect depends on decompression rate, melt inclusion size, and crystallographic direction. Our modeling approach shows that factors such as 3D crystal geometry and melt inclusion configuration can play an important role in constraining accurate decompression rates and recovering water contents in natural magmatic systems.