Numerical considerations in simulating the global magnetosphere

Abstract

Abstract. Magnetohydrodynamic (MHD) models of the global magnetosphere are very good research tools for investigating the topology and dynamics of the near-Earth space environment. While these models have obvious limitations in regions that are not well described by the MHD equations, they can typically be used (or are used) to investigate the majority of magnetosphere. Often, a secondary consideration is overlooked by researchers when utilizing global models – the effects of solving the MHD equations on a grid, instead of analytically. Any discretization unavoidably introduces numerical artifacts that affect the solution to various degrees. This paper investigates some of the consequences of the numerical schemes and grids that are used to solve the MHD equations in the global magnetosphere. Specifically, the University of Michigan's MHD code is used to investigate the role of grid resolution, numerical schemes, limiters, inner magnetospheric density boundary conditions, and the artificial lowering of the speed of light on the strength of the ionospheric cross polar cap potential and the build up of the ring current in the inner magnetosphere. It is concluded that even with a very good solver and the highest affordable grid resolution, the inner magnetosphere is not grid converged. Artificially reducing the speed of light reduces the numerical diffusion that helps to achieve better agreement with data. It is further concluded that many numerical effects work nonlinearly to complicate the interpretation of the physics within the magnetosphere, and so simulation results should be scrutinized very carefully before a physical interpretation of the results is made. Our conclusions are not limited to the Michigan MHD code, but apply to all MHD models due to the limitations of computational resources.