Improving CME evolution and arrival predictions with AMR and grid stretching in Icarus

Context. Coronal mass ejections (CMEs) are one of the main drivers of disturbances in interplanetary space. Strong CMEs, when directed towards the Earth, cause geomagnetic storms upon interacting with the Earth's magnetic field, and can cause significant damage to our planet and affect everyday life. As such, efficient space weather prediction tools are necessary to forecast the arrival and impact of CME eruptions. Recently, a new heliospheric model called Icarus was developed based on MPI-AMRVAC, which is a 3D ideal magnetohydrodynamics (MHD) model for the solar wind and CME propagation, and it introduces advanced numerical techniques to make the simulations more efficient. In this model the reference frame is chosen to be co-rotating with the Sun, and radial grid stretching together with adaptive mesh refinement (AMR) can be applied to the numerical domain. Aims. Grid stretching and AMR speed up simulation results and performance. Our aim is to combine the advanced techniques available in the Icarus model in order to obtain better results with fewer computational resources than with the equidistant grid. Different AMR strategies are suggested, depending on the purpose of the simulation. Methods. In this study, we model the CME event that occurred on July 12, 2012. A cone model was used to study the CME's evolution through the background solar wind, and its arrival at and impact with the Earth. Grid stretching and AMR were combined in the simulations by using multiple refinement criteria, to assess its influence on the simulations' accuracy and the required computational resources. We compare simulation results to the EUHFORIA model. Results. We applied different refinement criteria to investigate the potential of solution AMR for different applications. As a result, the simulations were sped up by a factor of ∼17 for the most optimal configuration in Icarus. For the cone CME model, we found that limiting the AMR to the region around the CME-driven shock yields the best results. The results modelled by the simulations with radial grid stretching and AMR level 4 are similar to the results provided by the original EUHFORIA and Icarus simulations with the 'standard' resolution and equidistant grids. The simulations with 5 AMR levels yielded better results than the simulations with an equidistant grid and standard resolution. Conclusions. Solution AMR is flexible and provides the user the freedom to modify and locally increase the grid resolution according to