» Back to overview
Proposing Institution

Institute for Meteorology and Climate Research
Project Manager

Dr. Dominikus Heinzeller
82467 Garmisch-Partenkirchen
The Model for Prediction Across Scales (MPAS) is a novel set of earth-system simulation components and consists of an atmospheric model, an ocean model and a land-ice model. Its distinct features are the use of unstructured Voronoi meshes and C-grid discretisation to address shortcomings of global models on regular grids and of limited area models nested in a forcing data set, with respect to parallel scalability, numerical accuracy and physical consistency.In a recent study using computational resources provided through a PRACE preparatory access proposal, we investigated the scalability of MPAS on various supercomputing facilities in Europe. We further performed an extreme scaling experiment on a 3 km mesh with more than 65 million grid cells on up to 458752 cores within the recent Juqueen Extreme Scaling Workshop 2015. We also conducted a first experiment to assess the physical accuracy of the MPAS-A atmospheric model using regular and variable-resolution meshes with a special focus on West Africa. We could demonstrate that MPAS is able to scale out on massively parallel systems and to correctly reproduce the atmospheric dynamics on global and local scales. This makes MPAS a promising tool for conducting climate-related impact studies of, for example, land use changes in a consistent approach.As a next step, we propose to explore the potential benefits of using a variable-resolution mesh with a special focus on modelling the African monsoon systems with a large, high-resolution area centered over West Africa and East Africa. For this, we plan to conduct a parameter study to optimise the configuration of MPAS for the region and compare the MPAS model performance to observational data sets and WRF reference runs obtained within the WASCAL program (http://www.wascal.org) and data from the CORDEX initiative (http://wcrp-cordex.ipsl.jussieu.fr).In preparation for this experiment, we here apply for a test account on SuperMUC in order to perform scaling experiments with the selected mesh to derive an optimal parallel setup and obtain reliable numbers on the required computational resources.

Impressum, Conny Wendler