The horizontal grid size in NWP models will continue to decrease. There are many reasons to decrease the horizontal grid size in such models. In general, as the grid size decreases, small-scale processes become better resolved so that the need to parameterize subgrid-scale processes also decreases.

In particular, as the grid size decreases, the processes that determine the location, timing, and intensity of convective rainfall should become better resolved, which should improve forecasts of convective precipitation. Triggering of convection will be better represented, due to increased resolution of mesoscale boundary-layer circulations. As the horizontal grid size decreases, many of the processes that determine the structure of mesoscale convective systems (MCSs) will also be better resolved.

However, until horizontal grid sizes decrease to 4 km or less, cumulus convection will still need to be parameterized. As horizontal grid sizes decrease, parameterized convection becomes more localized. That is, a smaller fraction of the grid columns are convective columns (i.e., columns that contain areas of cumulus convection) and these columns include larger fractional areas of cumulus convection. What are the implications for cumulus parameterization as horizontal grid size decreases and convection becomes more localized? Can we quantify the implications? Do some aspects of the cumulus parameterization for large-scale models become inappropriate as horizontal grid size decreases? If so, which ones?

We used the results of a 29-day, 2D CRM simulation of mostly deep convective cloud systems over Oklahoma during the summer to quantify the dependence of some aspects of cumulus parameterization on horizontal grid size. The CRM's horizontal grid size was 2 km, so that it resolved both the cloud-scale and the mesoscale vertical motions associated with deep convective cloud systems.

We identified the convective grid columns in the CRM simulation using the method of Xu (1995). The convective grid columns are the convective portions of the simulated deep convective cloud systems. A grid cell refers to a hypothetical grid column in a mesoscale or large-scale model; we considered grid cells with horizontal grid sizes that range from 2 km to 256 km. A convective grid cell contains at least one convective grid column.

We have analyzed the dependence of the following quantities, among others, on horizontal grid size: fraction of grid cells that are convective grid cells, resolved convective mass flux, resolved non-convective mass flux, resolved fraction of convective mass flux, and average convective column fraction in the convective grid cells. These results and our conclusions will be presented at the conference.