This title appears in the Scientific Report : 2018 
Adaptive Monte Carlo sampling for cloud and microphysics calculations
Rössler, Thomas (Corresponding author)
Jülich Supercomputing Center; JSC
2017
ix, 53 p.
Masterarbeit, The University of Wisconsin{Milwaukee, 2017
Master Thesis
Computational Science and Mathematical Methods
OpenAccess
OpenAccess
Please use the identifier: http://hdl.handle.net/2128/19838 in citations.


An important problem in large-scale modeling of the atmosphere is the parametrization ofclouds and microphysics on subgrid scales. The framework Cloud Layers Unified By Binormals(CLUBB) was developed to improve the parametrization of subgrid variability. MonteCarlo sampling is used to couple the different physical processes, which improves the gridaverage of subgrid tendencies. In this thesis we develop an adaptive Monte Carlo samplingalgorithm that re-uses sample points of the previous time step by re-weighting them accordingto the change of the underlying distribution. This process is called "what-if sampling"and is an application of importance sampling. An example illustrates that what-if samplingconverges slowly when the atmospheric conditions change too much. Therefore, the algorithmwas extended by adaptive criteria. These prohibit re-weighting if the atmosphericconditions change too fast and allow the re-weighting method to converge to the right solution.We studied five test cases for different atmospheric conditions and found that thecomputation of the what-if weights is too expensive and suffers from bad importance sampling.The high-dimensional distribution of CLUBB that evolves in time makes re-weightingdifficult. The simulation results of what-if sampling are similar to the standard Monte Carlomethod or even worse considering the higher computational costs. Therefore, the algorithmwas simplified such that old tendencies are re-used without any re-weighting. This approximationremoves the overhead and reduces the extra noise. However, simple re-using doesnot improve the accuracy of the model for the same computation time for general appliiication. Only in the special case of very few sample points, this method can improve theperformance without increasing the error significantly. The standard Monte Carlo samplerof CLUBB works very efficiently by applying well suited importance sampling. For normalsimulations, using fewer sample points is better than applying any re-using algorithm to alarger number of sample points.