A key issue in modern numerical weather prediction arises in models with grid length ~ 10km, and concerns the parameterization of convection and its mesoscale organization. At this resolution the scale-separation assumption that underpins convective parameterization is not strictly valid. Moreover, convection may be explicit but under-resolved as well as being represented in a parameterization scheme. This distortion or surrogate behavior is an unphysical manifestation of convective organization that leads to model error and statistical bias. While this '10-km barrier' is universal, addressing the issue over the US continent has obvious advantages.
We examine the U.S. warm-season convection problem using the explicit (cloud-system-resolving modeling) approach. This model has been validated extensively validated against field experiment results. The talk will describe: i) Issues arising regarding the prediction of sequences of wasrm-season convection over the US contintent in regional weatehr prediction models; ii) the physical basis of these issues; iii) results obtained from cloud-system-resolving models, and iv) a physically based strategy for the parameterization of organized convection.