Operational global numerical weather prediction models already have a grid-spacing of about 20 km. Experimental global climate models have been run at grid-spacing less than 10 km. Next-generation computers will enable climate models to run century-long integrations at 10km grid-spacing. It follows that issues (e.g., the nature of systematic errors) arising with intermediate 10-km-grid models will be highly relevant in the foreseeable future.

The physical resolution of a numerical model is 7-10 times coarser than the grid spacing. Hence cumulus elements embedded in organized convection have to be parameterized in 10 km-grid models, but the classical scale-separation assumption is not strictly valid. The mesoscale organization of precipitating convection is represented but is under-resolved by explicit (grid-scale) circulations. Hence representing deep convection and its organization in global models rests at an interesting juncture. This kind of problem is addressed for warm-season convection over the continental US, although it has a general application.

A key issue is how to represent the dynamical mechanisms that are conspicuously absent from contemporary convective parameterizations. A hierarchical modeling strategy is adopted: cloud-system-resolving model simulations (3 km grid-spacing) and 10-km-grid simulations. Numerically simulated convective systems are physically interpreted using a mechanistic dynamical model of organized convection -- a formal basis for approximating mesoscale convective organization (stratiform heating and mesoscale downdrafts) by an analytic first-baroclinic couplet. Both convective heating and convective momentum transport are represented. In other words, the new hybrid parameterization consists of a convective parameterization (cumulus elements) and explicit convectively-driven circulations (convective organization). However, the under-resolved explicit circulations create specific systematic errors (e.g., too weak stratform heating and mesoscale downdrafts). This aspect is addressed by a predictor-corrector strategy: explicit dynamics is the predictor and the first-baroclinic couplet the corrector. The corrector strengthens the systematically weak mesoscale downdrafts at 10-km grid spacing.

When applied to the Betts-Miller-Janjic convective parameterization, this new hybrid approach emulates the propagation, hydraulic dynamics of organized precipitating systems. Therefore, the predictor-corrector approach is a potential framework for representing organized convection in models of intermediate resolution.