Background Science Applications Physics Climate Groundwater Fusion Energy Life Sciences Materials & Chemistry SciDAC Institutes Collaboratories Enabling Technologies Applied Mathematics Computer Science Visualization & Data Mgt. SciDAC Outreach Participating Orgs Grant Solicitations FY2007 FY2006 FY2005 FY2004 FY2001 Collateral Materials SciDAC Review magazine '06 Progress Report (pdf) Publications 2001-5 (pdf) Logos Search Collateral (from SciDAC 1) Browse by Affiliation Browse by Author Browse by Title DOE Office of Advanced Scientific Computing Research (ASCR) Program Managers only

Collaborative Design and Development of the Community Climate System Model for Terascale Computers

PI: R. Malone⁴, J. Drake⁵ Co-PIs: C. Ding², S. Ghan⁶, D. Rotman³, J. Taylor¹, J. Kiehl⁷, W. Washington⁷, S.-J. Lin⁸, ISIC affiliates: R. Armstrong (CCTTSS), D. Bailey (PERC), P. Colella (APDEC), J. Glimm (TSTT), D. Keyes (TOPS), Collaboratory affiliates: D. Middleton (ESG), C. Deluca (NASA-ESMF)

¹Argonne National Lab, ²Lawrence Berkeley National Lab, ³Lawrence Livermore National Lab, ⁴Los Alamos National Lab, ⁵Oak Ridge National Lab, ⁶Pacific Northwest National Lab, ⁷National Center for Atmospheric Research, ⁸NASA-Goddard Space Flight Center

Summary

The purpose of this multi-disciplinary project is to accelerate the development of the Community Climate System Model (CCSM). An early focus on software design and engineering of the CCSM and its component model has lead to performance improvements of the dynamical cores and other critical aspects of the atmosphere, ocean, land, and sea ice models. The introduction of atmospheric chemistry and ocean biogeochemistry into the model suite along with new formulations and algorithms for ocean and ice dynamics are extending the ability of the model to simulate complex climate interactions and feedbacks. Improving the science in the model as well as providing efficient parallel implementations supports the DOE’s Climate Change Prediction Program. Active participation in the open design process of the CCSM working groups, collaboration with key NASA and NSF research centers and cooperation with the Earth System Modeling Framework are coordinated with the CCSM Scientific Steering Committee.

The first full year of the project shows rapid progress on several tasks. Project teams organized to contribute to working groups in software engineering, ocean, atmosphere, land, chemistry and biogeochemistry. The first steps toward a CCSM model following an "open" design have now been succeeded by the second and third steps with the release of the CCSM2 in May 2002 and the establishment of a Code Review Board for atmospheric model software.

Progress in code development, in accordance with proposed designs of the atmosphere and coupler components, has resulted in significant performance improvements of the CCSM2. A new version of the land model CLM2.1 has been incorporated that includes new modular data structures and optimized parallel algorithms. The incorporation of the projects coupler utilities, MPH and MCT are evident in the new coupler, CPL6, which is being tested for deployment with the full model. The release of two utility libraries, MCT1.0 and the ZioLib show a new level of code maturity and robust performance.

Other significant progress includes the optimization of the 2-D decomposition of the Lin-Rood (finite volume) dynamical core, the introduction of new parallel spectral decompositions in the Eulerian Spectral and semi-Lagrangian Spectral dynamical cores and scalability improvements from load balanced parallel decompositions of the "chunked" physics. The load balancing has become even more important as the model is exercised with atmospheric chemistry options.

Substantial progress has also been made in several aspects of the ocean and sea ice models. A new version of POP, version 2.0, has been completed and will be released soon. It differs from POP1.4.3, the last formal release, in a number of ways. Partial bottom cells and the new tripole grid have been implemented in POP2.0 expanding the model options. Small computational domains that can be contained in cache or much larger domains that are compatible with vectorization, allows POP to perform well on cache or vector machines like the Japanese Earth Simulator and the new Cray X1.

Testing of a new ocean model formulation in the hybrid vertical-coordinate version of POP, called HYPOP, reached a new stage. Tests emphasized the two extreme limits of the hybrid coordinate, fully Lagrangian and fully Eulerian. It was found that the two solutions match closely, but, as expected, the Eulerian limit is more diffusive.

Incorporation of tropospheric chemistry moved from the design phase to the implementation phase with multi-year global simulations and the first diagnostic look at greenhouse gases advected by the model. A configuration of the model using the FV core and the WACCM stratospheric chemistry package was also considered.

The first steps were made on ocean biogeochemistry, both organizationally and technically. Studies of nutrient upwelling and chlorophyll production, the effects of iron enrichment, and the world-wide distribution of dimethyl sulfide (DMS) were performed using this new capability.

Meetings with the Evaluation Research Center (PERC) and the Earth System Grid (ESG) Collaboratory were facilitated by the Access Grid. These joint meetings were quite helpful. A Memorandum of Understanding (MOU) with the NASA Earth System Modeling Framework (ESMF) was put in place and representatives from our project have become active participants in ESMF meetings, design review and code development. A meeting with the SciDAC Common Component Architecture project and ESMF resulted in the definition of tasks for the MCT utility and for the atmospheric model software development.

Figure 1. Surface chlorophyll distributions simulated with the biogeochemical version of the Parallel Ocean Program (POP) for conditions in late 1996 (a La Niña year).

The SciDAC CCSM project webpage may be found at http://www.scidac.org and the full CCSM project pages may be found at http://www.ccsm.ucar.edu.

For further information on this subject contact:
John Drake (drakejb@ornl.gov) or Bob Malone (rcm@lanl.gov).
Phone: 865-574-8670 or 505-660-3663

back to project page