An idealized comparison of one-way and two-way grid nesting

Most mesoscale models can be run with either one-way ("parasitic") or two-way ("interactive") grid nesting. We present results from a linear 1D shallow-water model and a 3D simulation of a multicell thunderstorm to determine whether the choice of nesting method can have a significant impact on the solution.

Even for well resolved disturbances on the coarse grid, it is found that one-way nesting produces significantly more reflection than two-way nesting when a disturbance attempts to exit the nested grid. The reflections that do occur when using two-way nesting can be reduced to negligible levels with a Davies "relaxation" nested boundary condition. Reflections in the one-way case are attributed to the accumulated phase errors between the solutions on the two grids, leading to the wrong boundary data being supplied to the nested grid and the creation of a spurious incoming mode. In the one-dimensional shallow water simulations, this can cause reflections nearly twice as large in amplitude than the incident wave, while the multicell simulations show unphysical errors in precipitation of up to 20% locally. This reduction in reflection amplitude when using two-way instead of one-way nesting is found to be robust for a range of modifications to the relaxation BC.