Computing Cloud Dynamics with an Adaptive Monte Carlo Sampling Method

Thomas Roessler
University of Wisconsin-Milwuakee
Graduate Student

“Atmospheric models usually have simple parametrizations of subgrid variability that do not include all processes. The model Cloud Layers Unified By Binormals (CLUBB) uses a joint multivariate normal distribution to model clouds and applies Monte Carlo sampling to compute more accurate time tendencies for fluxes on the subgrid scale. However, this requires the expensive computation of the microphysics for each sample point. We will apply a re-weighting method called ‘what-if sampling’ to re-use sample points of previous time steps. The number of re-uses is controlled adaptively to minimize the number of evaluations of the microphysics by maintaining the simulation quality”.

View Roessler Event Flyer