Reducing Computational Cost of Convective System LES via Translating Nests
Understanding the dynamics of large convective systems increasingly requires high-resolution simulations to obtain small-scale details within the mesoscale and synoptic contexts. Large-eddy simulation (LES) is a powerful tool for obtaining these details. However, organized convective systems exist over hundreds of kilometers and propagate over large regions. This makes traditional LES approaches untenable because a domain containing the system would be too large and expensive. Alternatively, one can use a smaller domain that translates with the convective system instead of a static, large domain. This reduces cost by removing grid columns that have little impact on the storm development and instead focuses computation on the area of interest as it moves over time. While this approach is regularly used for hurricanes, it has not been documented for mesoscale convective systems.
We applied the translating domain methodology to simulate the 20 May 2011 mesoscale convective system that occurred during the Midlatitude Continental Convective Clouds Experiment (MC3E) over Oklahoma. A benchmark simulation with a large, static domain is used to compare with the smaller translating domain. We find that while the convection is not identical between the two simulations, they are very close to each other with very few numerical artifacts due to the domain movement. The convective details are sufficiently close that the translating domain provides a suitable dataset for convective studies. The benchmark vs. small-domain comparison was done with 400 m grid spacing to manage computational cost. An additional simulation using the translating nest has been done with 240 m grid spacing to reach LES grid scales, providing a powerful dataset for understanding how convective cells change as the storm transitions from initiation, upscale growth through linear organization, and decay into smaller organized units over an 18 h period.