The Influence of Mesoscale, Under-Resolved Convective Structures on Nested Large Eddy Simulations

Accurate, multi-scale models of atmospheric phenomena are essential for making predictions of weather conditions, wind energy production, pollution transport and fire propagation. These models must take into account mesoscale atmospheric flow patterns while also representing the microscale flow characteristics. Simultaneously representing these two scales often requires bridging between mesoscale and microscale models. In our study we quantify the challenges of applying the grid-nesting technique for the purpose of coupling mesoscale and large eddy simulations (LES), in the presence of atmospheric convective structures.

Atmospheric convection provides unique challenges due to the length scale of convective motions. The size of atmospheric convective structures falls within a region of scales known as the terra incognita (TI), which neither mesoscale nor microscale models have been designed to represent. Unfortunately, due to grid-nesting limitations, when nesting between a mesoscale parent domain and a microscale domain, the mesoscale simulation is usually run at resolutions within the TI. The result of this is the presence of unrealistic, grid-size dependent flow structures in the parent domain that are transferred to the nested domain through boundary conditions.

In this study we perform two sets of nested mesoscale-to-LES runs using the Weather Research and Forecasting Model (WRF) for different convective structures: rolls and cells. We use resolutions within the TI for the parent mesoscale domain and study the effect of the unrealistic structures on the nested LES domain. By comparing the large-scale and turbulent characteristics of the nested LES domain to those of a stand-alone, periodic LES domain we are able to quantify the influence of mesoscale flow structures. This new understanding of the limitations of grid-nesting in the context of mesoscale-to-LES will improve nesting practices and inform developers as to where model improvements are needed most.