Over the past forty years, the numerical simulation of atmospheric convection has evolved from its infancy in two-dimensional dry thermals to highly sophisticated three-dimensional models used for numerical weather prediction at convective scales. This advancement has been feasible because of the enormous growth in computing power, together with significant advancements in model numerics, physical parameterizations, and data analysis required to capture the complexity of convective severe weather in numerical simulation models. Examples of recent advances in the numerical simulation of convective weather include the assimilation of Doppler radar observations that provide high-resolution data for model initialization, two moment microphysical parameterizations that better differentiate the precipitation types and their size distributions, and moving nested grids that allow the selective placement of high resolution in regions of significant convective activity. Numerous operational mesoscale forecast models are now moving to horizontal grids on the order of several kilometers in which convective severe weather is treated explicitly without the use of cumulus parameterization.
In spite of the rapid progress in numerically simulating convective systems, significant challenges remain in seeking to advance the numerical prediction of convective-scale weather, particularly in the areas of data assimilation and model physics. Because of the important small-scale structure in convective storms, optimal assimilation techniques must be developed for all available data, including both ground-based satellite observations. Both advanced variational data assimilation approaches and ensemble Kalman filter techniques should be explored in seeking a complete and dynamically consistent of the initial atmospheric structure.
Given accurate initial conditions, the numerical model must also accurately capture the important cloud- and mesoscale processes that control the evolution of convective severe weather. The uncertainties and approximations in the physics parameterizations are believed to be the most significant contributors to model error. The cloud microphysics, in particular, contain significant sources of uncertainty for explicit prediction of convective cells. Multi-species microphysics schemes with more accurate particle size distribution models and/or multiple moment schemes should be refined and verified against observations for different types of storms. The planetary boundary layer (PBL) also plays a strong role in the development of convective systems; new PBL schemes are required that are suitable for kilometer-scale resolutions where a significant portion of convective boundary layer mixing is achieved by resolvable eddies. Subgrid-scale turbulence closure models suitable for non-LES resolutions also require further research.
As numerical models improve their ability to simulate convective weather events, further research is needed to understand their limits of predictability. Because of the inherently short time and space scales of convection, probabilistic techniques using ensemble approaches should be explored. New techniques also need to be developed to quantitatively assess the accuracy and value of numerical forecasts produced by these cloud-resolving models.
We invite submissions of contributions on all the above aspects.