Abstract
Numerical models for atmospheric dynamics, a central piece in weather and climate forecasting, are going through radical changes in order to become adequate for the next generation supercomputers that demand massively parallel processing.This project aims to study two different lines of numerical schemes for this next generation of models: (i) the development and analysis of numerical schemes for quasi-uniform spherical grids, that avoids scalability problems encountered in traditional models; (ii) the exploration of the time dimension as source of parallelism, with methods that radically change the time integration way of thinking of these models.Recently, we have shown that one of the main methodologies used in this new generation of quasi-uniform grid based models lacks consistency (local truncation error fails to converge to zero with increasing resolution). We propose a scheme that circumvents this issue and we aim to add this new scheme to the National Center for Atmospheric Research (NCAR-USA) Model for Prediction Across Scales (MPAS), envisioning to build a numerically and dynamically consistent version of this model. In parallel, we will study other problems arising with this kind of approach, such as numerical instabilities and ways to optimize locally refined quasi-uniform grids. The usage of quasi-uniform grids extends the utility of atmospheric models with respect to its adequacy to modern computer architectures, however, in the long run, their scalability will still be limited by its domain partition communication pattern. We aim in this project to explore the time dimension as a new source of parallelism. Recently, we developed a massively parallel exponential time integrator method, that we intend to couple with a semi-Lagrangian integrator, for nonlinear advection, and with an iterative parallel-in-time scheme, to build a pioneer parallel-in-time method adequate for atmospheric models.This project will be accomplished mainly with the collaboration of researchers from UK universities (Univ. of Exeter and Imperial College), American research centres (Los Alamos Nat. Lab. and NCAR) and the University of Sao Paulo. (AU)
| Articles published in Agência FAPESP Newsletter about the research grant: |
| TITULO |
| Articles published in other media outlets (0 total): |
| More itemsLess items |
| VEICULO: TITULO (DATA) |
| VEICULO: TITULO (DATA) |