Project proposes a three-year research and development effort, aimed at enabling a broad climate science capability for petascale systems, with four major thrusts: ? Investigating the suitability of next-generation computer languages and compilers in prototype petascale coupler component design. ? Extending the Flux Coupler design to efficiently support embedded ensemble based climate modeling techniques. ? Providing rich opportunities for interdisciplinary research and education by bringing computer scientists and climate researchers and their graduate students together to work on this project. ? Working with the CCSM core group and others to ensure the fullest possible dissemination of our software and results.
Investigators will use novel interactive ensemble and stochastic physics techniques to obtain new insights about many basic but poorly understood questions in climate dynamics that depend on the stochastic nature of fluctuations in the Earth?s atmosphere, oceans, cryosphere, and land surface. Interactive ensembles embedded in climate models will enable the non-trivial use of large processor counts within a single coupled climate application. The flexibility and power of PGAS languages will be used address the scalability and complexity challenges of the petascale Flux Coupler.
Broader Impacts
The CCSM is a community model used by hundreds of researchers, and is one of the climate models used in the International Panel on Climate Change (IPCC) assessments. By enabling the CCSM to make use of petascale systems, this project will permit the U.S. to keep pace with similar efforts in Europe and Japan. This research will provide a basis for improving our understanding of the role of noise in climate system dynamics for seasonal prediction and global climate change. While this proposal focuses on climate science, the byproducts of the work are applicable to coupled modeling problems in other science and engineering fields, particularly the geosciences, and can inform the long-range design plans of other coupling tools and frameworks. Finally, the project will bring together graduate students in computer science and climate science to work on petascale problems, thus training the next generation of computational scientists.
