Strategies for Performing Large-Eddy-Simulations in Weather Models without Paying the Price
Please login to view abstract download link
We have developed a multi-modeling framework (MMF) designed to replace the physical parameterizations used in weather/climate models. We began this work by viewing the coarse-scale and fine-scale models through the lens of Variational Multi-Scale (VMS) methods in order give MMF a more rigorous mathematical foundation. We continue along this path but have evolved our thinking to see the connection between MMF and block-structure adaptive mesh refinement. Although MMF offers large savings in time-to-solution over a standard global fine-scale simulation (at the large-eddy-simulation scale), MMF is still more expensive than standard parameterizations. To circumvent this cost we explore GPU computing and reduce-order modeling approaches including neural networks. We present results for three-dimensional simulations of moist supercells.
