Three microphysics schemes are available: single-moment and double-moment bulk schemes and the Lagrangian super-droplet method (SDM). The SDM is the most detailed, but also most computationally demanding of the three. It can be calculated on a CPU using OpenMP or on CUDA GPUs. If the latter option is used, fluid flow calculations can be done by CPU in parallel with microphysics calculations done by GPUs. The SDM implementation allows users to choose from a number of coalescence kernels, terminal velocity formulas and initialization schemes. It also allows for modeling chemical reactions within droplets and aerosol sources.
The model can make use of a number of solvers from the libmpdata++ library, most notably from the third-order accurate variant of the MPDATA scheme. Regarding subgrid-scale models, currently the Smagorinsky model is available. The user can choose to perform implicit LES simulations, for which the MPDATA scheme is well suited.
The model uses domain-decomposition on distributed memory systems. It makes use of the recent MPI implementations that are CUDA-aware and thread-safe. For maximum performance, intra-node theading is done using OpenMP.
Results of modeling stratocumulus and cumulus clouds using UWLCM will be compared with other models.