|Position Title||Petascale Numerical Weather Prediction with the Weather Research and Forecasting (WRF) Model|
|Summary||The pre-selected student, Kayla Harrison, will work in a year-long project to push a current state-of-the-art numerical weather prediction model into petascale environments for high resolution modeling in complex terrains. The study will focus on understanding both the computational issues of working with grand-scale model domains, and the numerical issues of modeling at very fine resolutions in topography that is typically troublesome for modellers.|
|Job Description||This project involves research aimed at setting the stage for the next-generation weather models, capable of successful operation at horizontal resolutions of hundreds to tens of meters in real-world regimes. The model domains we are interested in are those that are most challenging - huge, high resolution domains in regions of complex topography. This work requires a two-pronged emphasis that focuses both on the computational resources required for such high-resolution simulations, and the sensitive model physics that often break down at such high resolution.|
The computational resources aspect has been explored by myself and others as we have pushed the Weather and Research Forecasting (WRF) model to domain sizes of over two billion grid points. These types of simulations tend to push the technology and require innovative methods to get them running at the large scale. This benchmarking work is, additionally, being pursued through collaborations under the leadership of scientists at the Institute of Meteorology at Vienna’s University of Natural Resources and Life Sciences, and the Institute of Energy Technologies at the Technical University of Catalonia in Barcelona. In this work, we are exploring the use of nested grids at sub-km resolution over the Alps - in a real-world modelling environment - in an effort to benchmark both computational and numerical performance in modeling domains of interest to the typical atmospheric modelers.
The goals of this particular year-long project will be to produce a publishable report that discusses the current limits of high-resolution numerical weather prediction (with WRF) from the perspectives of computational limits and numerical stability. The study will investigate the problems and limits of pushing WRF to very large scales, including issues of scalability, domain nesting and I/O, which becomes prohibitive at large scales. Simultaneously, the study - through the availability of petascale computing resources will investigate the numerical effects of large, high-resolution grid scales, in an effort to find ways to minimize some of the instabilities through correct domain configurations and choice of model parameterizations.
The student recommended for this position, Kayla Harrison, has worked with this mentor at the University of Alaska Fairbanks (UAF) Arctic Region Supercomputing Center (ARSC) for two years on issues related to numerical weather prediction, primarily using the Weather Research and Forecasting (WRF) model, and miscellaneous auxiliary postprocessing tools (VAPOR, NCL and Python Scripts). She has also been collaborating with our European collaborators to produce a poster presented at the 2011 Alaska Weather Symposium and accepted for presentation at the 2011 International Conference on Alpine Meteorology.
|Location||Arctic Region Supercomputing Center, University of Alaska Fairbanks, Fairbanks, Alaska|