If you suspend your transcription on amara.org, please add a timestamp below to indicate how far you progressed! This will help others to resume your work!
Please do not press “publish” on amara.org to save your progress, use “save draft” instead. Only press “publish” when you're done with quality control.
Today the evidence for global climate change is unequivocal, and the human influence is clear. Therefore the focus of young researchers has shifted from assessing whether the Planet is warming towards envisioning how a warmer world might look like. For instance, basic physical principles suggest that the hydrological cycle of Planet Earth will likely undergo dramatic changes. However, understanding and describing the involved processes, estimating future changes, and assessing the underlying uncertainties has proven to be difficult and complex. In this effort, numerical simulations of the weather and climate system are a useful research tool.
Weather and climate modeling involves solving the governing equations of atmospheric motion on a numerical mesh and employing semi-empirical parameterizations that treat the processes not represented explicitly. For example, the parameterizations typically include treatments for thunderstorms and rain showers (deep convection). These processes are fundamental to the climate system since they vertically redistribute moisture, heat, and momentum, but so far they could not be resolved explicitly, due to the coarse gird spacing of the mesh (resolution) employed in the current generation of climate models.
In the recent year's power constrains in the domain of supercomputing have lead to heterogeneous node designs mixing conventional multi-core processors and accelerators such as graphics processing units (GPU’s). These machines posses properties beneficial for weather and climate codes and hence allow refining the resolution of the involved computational mesh to the kilometer scale. Convective clouds can then be represented explicitly (convection-resolving) and the models can be formulated much closer to physical first principles. However, to exploit the capabilities of these supercomputers, model codes have to be ported, a challenging task the weather and climate modeling community is struggling with.
We discuss prospects and challenges climate modelers face on these new supercomputers and highlight the potential for addressing key open science questions. The presentation is illustrated with simulations recently accomplished using a new version of the Consortium for Small-Scale Modeling weather and climate model (COSMO), capable of exploiting these heterogeneous supercomputer architectures. Using results form a then-year-long climate simulation on a computational domain covering Europe (1536x1536x60 grid points) we highlight some of the added value of the approach regarding the representation of precipitation processes. Furthermore, we explore the gap between the currently established regional simulations and global simulations by scaling the GPU accelerated version of the COSMO model to a near-global computational domain.
References:
Fuhrer, O., Chadha, T., Hoefler, T., Kwasniewski, G., Lapillonne, X., Leutwyler, D., Lüthi, D., Osuna, C., Schär, C., Schulthess, T. C., and Vogt, H.: Near-global climate simulation at 1 km resolution: establishing a performance baseline on 4888 GPUs with COSMO 5.0, Geosci. Model Dev. Discuss., https://doi.org/10.5194/gmd-2017-230, in review, 2017.
Leutwyler, D., Lüthi, D., Ban, N., Fuhrer, O., and Schär, C.: Evaluation of the Convection-Resolving Climate Modeling Approach on Continental Scales, J. Geophys. Res. Atmos., 122, doi:10.1002/2016JD026013
Leutwyler, D., Fuhrer, O., Lapillonne, X., Lüthi, D., and Schär, C., 2016: Towards European-scale convection-resolving climate simulations with GPUs: a study with COSMO 4.19, Geosci. Model Dev., 9, 3393-3412, doi:10.5194/gmd-9-3393-2016.