In an effort to strike balance between the fidelity of direct numerical simulations and the computational efficiency of continuum models of soils, a hierarchical multiscale model was developed to model the large strain, high strain rate split Hopkinson pressure bar experiments conducted on dry Colorado Mason sand. Two, one-dimensional continuum methods, the finite element and material point methods (FEM and MPM), were used to model the geometry of the experiment, and rather than developing a phenomenological constitutive model for the dry sand, ellipsoidal discrete element method (DEM) particle assemblies were generated from synchrotron microcomputed tomography imaging of the the soil and used in place of the constitutive model in the continuum methods. The continuum methods calculated the deformation gradient over the domain by solving the momentum equation, which was then passed as boundary conditions to the DEM assemblies. After deforming, a homogenized Cauchy stress over the DEM assemblies was passed back to the continuum models. Due to evidence of particle crushing in the experimental data, a particle fracture model was added to the discrete element method code, and the full code was implemented using a hybrid OpenMP/MPI parallelization scheme for use on supercomputers to significantly reduce the overall computational expense. Results from the simulations show the ability to reasonably reproduce the experimental results with the inclusion of the particle fracture model, and the novel hierarchical multiscale MPM-DEM model was introduced and compared to the FEM-DEM simulation results.