2.2      Results from the implementation of an implicit-explicit vertical advection scheme in WRF-ARW for CAM systems.

Wicker, Lou, NOAA/NSSL

Over the last two decades computational power has increased sufficiently to enable prediction of high-impact weather using Convection Allowing Models (CAMs) at nearly every major numerical weather prediction center. Due to their high cost, CAM models almost always use horizontal and vertical grid resolutions having large aspect ratios (△x  1). Consequently the CAM’s △z time step is often constrained by the vertical Courant-Levy-Friedrichs (CFL)
number associated with strong convective drafts which typically occur only in a few grid columns. To increase operational efficiency, CAMs often employ extra filters to reduce maximum updraft speeds that then permits the use of a larger time step. In this case, integration efficiency is increased at the
expense of solution accuracy. Alternatively, implicit advection methods often have no time step restrictions, but introduce much larger phase and amplitude
errors as CFL exceeds  1. A hybrid vertical advection method currently used in a three-dimensional ocean model is implemented in NCAR’s Weather
Research and Forecasting model to improve the efficiency of the model. The method combines explicit and implicit advection methods to a create a scheme
enabling stable integrations even when the vertical advection CFL number exceeds 2. Tests show that the scheme permits a ∼ 20 percent larger time step
 even with the removal of the currently used vertical velocity filters. Storm up-draft cores then retain most of their unfiltered intensities while still achieving
a ∼ 15 percent reduction in wallclock time for a full-physics CAM integration when compared to the filtered operational run. Several examples where the
IEVA demonstrates improved numerical efficiency will be shown.