Browsing by Subject "Adaptive Mesh Refinement"

Simulating space plasma on a global scale is computationally demanding due to the system size involved. Modeling regions with variable resolution depending on physical behavior can save computational resources without compromising too much on simulation accuracy. This thesis examines adaptive mesh refinement as a method of optimizing Vlasiator, a global hybrid-Vlasov plasma simulation. Behavior of plasma near the Earth's magnetosphere and different characteristic scales that need to be considered in simulation are introduced. Kinetic models using statistical methods and fluid methods are examined. Modeling electrons kinetically requires resolutions orders of magnitude finer than ions, so in Vlasiator ions are modeled kinetically and electrons as a fluid. This allows for lighter simulation while preserving some kinetic effects. Mesh refinement used in Vlasiator is introduced as a method to save memory and computational work. Due to the structure of the magnetosphere, resolution isn't uniform in the simulation domain, with particularly the tail regions and magnetopause having rapid spatial changes compared to the relatively uniform solar wind. The region to refine is parametrized and static throughout a simulation run. Adaptive mesh refinement based on the simulation data is introduced as an evolution of this method. This provides several benefits: more rigorous optimization of refinement regions, easier reparametrization for different conditions, following dynamic structures and saving computation time in initialization. Refinement is done based on two indices measuring the spatial rate of change of relevant variables and reconnection respectively. The grid is re-refined at set intervals as the simulation runs. Tests similar to production runs show adaptive refinement to be an efficient replacement for static refinement. Refinement parameters produce results similar to the static method, while giving somewhat different refinement regions. Performance is in line with static refinement, and refinement overhead is minor. Further avenues of development are presented, including dynamic refinement intervals.