The North Carolina Supercomputing Center (NCSC) will work with the National Center for Atmospheric Research (NCAR) and with the Lawrence Livermore National Laboratory (LLNL) to prototype the use of parallel-capable, self-describing data models in coupling the components of a climate model. This work would build on the disciplinary expertise of NCAR and LLNL in climate modeling and parallel computing, and on preliminary efforts at LLNL and NCSC using object-oriented technologies to build interdisciplinary modeling frameworks. Using these approaches can be a very powerful means of reducing simulation model complexity. Coupling the atmospheric, ocean, and ice components of a climate models is an excellent application. Present coupling strategies are fairly complex and somewhat ad hoc. Using appropriate abstractions, one can separate the science, the algorithms used for coupling, and the parallel data and task decompositions more cleanly, resulting in a more portable, maintainable, extensible, and comprehensible model. It is believed this can be done without unduly sacrificing efficiency. This award is made under the USGCRP Methods and Models for Integrated Assessments.
