The Customizable Unstructured Mesh Generation (CUMG) enables local mesh refinement in arbitrary shape using user-defined horizontal resolution in any desired locations. Using Hierarchical Time-Stepping (HTS), the model run time is reduced tremendously for meshes with large resolution variation. This study compares the mesh quality, real simulation efficiency, as well as predictive performance of the CPAS customized 128-to-1 km mesh to the MPAS-A standard 60-to-3 km mesh.
CPAS 128-to-5 km mesh contains 43,580 horizontal grid cells. It includes 5 regions covering ‘Asia’, ‘Himalaya’, ‘Southern China’, ‘Guangdong’ and ‘HongKong’ where the resolution transits from 128 km (outside) to 5 km (Hong Kong) with the ‘Maximum Resolution Gradient’ equals to 0.0833 km/km.
CPAS 128-to-1 km mesh contains 343,226 horizontal grid cells. It includes 5 regions covering ‘Asia’, ‘Himalaya’, ‘Southern China’, ‘Guangdong’ and ‘HongKong’ where the resolution transits from 128 km (outside) to 1 km (Hong Kong) with the ‘Maximum Resolution Gradient’ equals to 0.0833 km/km.
MPAS-A uses a global timestep (determined by the CFL condition on the smallest mesh cell) that substantially limits the resolution variation. The hierarchical time-stepping (HTS) feature implemented in ClusterTech Platform for Atmospheric Simulation (CPAS) relaxes this restriction by applying different time-steps to mesh cells of different sizes. This new feature enables extreme variable-resolution meshes and higher computational efficiency. This study evaluates the model results using HTS.
The ClusterTech Platform for Atmospheric Simulation (CPAS) is a cloud-based service platform which implements customized mesh generation and hierarchical time-stepping on Model for Prediction Across Scales - Atmosphere (MPAS-A (v6.3)) to better serve the computational needs of numerical atmospheric model users.