The Community Climate System Model, is a collaborative effort involving NSF, DOE and NASA aimed at providing US researchers with state of the art coupled climate simulation capabilities. The DOE national laboratories involved are ANL, LANL, LLNL, LBL, ORNL and PNL. The National Center for Atmospheric Research and the NASA/Goddard Data Assimilation Office are the primary, non-DOE collaborators. Phil Jones (pwjones@lanl.gov) and John Drake (drakejb@ornl.gov) are co-PI's for the project and the primary contacts.

Executive Summary

The DOE/SciDAC Accelerated Climate Prediction Initiative (ACPI) is focused on enhancing the nation's climate simulation capability. One strategy is to combine the capabilities of a major climate modeling institution, such as NCAR or GFDL, with those of DOE multi-disciplinary laboratories to speed the development of new generations of coupled climate models and component models. These new models must be physically comprehensive, utilize the most advanced algorithms and software engineering practices, and be portable with high performance across the latest computing platforms including highly parallel Terascale computers. In fiscal year 2000, DOE approved two pilot projects lasting 18 months. One of these projects, dubbed the Avant Garde(AG) project and led by Ian Foster (ANL), combined the efforts of NCAR, NASA/Data Assimilation Office (DAO), and five DOE laboratories to improve the Community Climate System Model (CCSM). Attention was focused on the atmospheric component (CCM3) and the coupler, and the scope of the project was restricted to issues of software and toolkit design and performance enhancement of the atmospheric dynamical core. During that time, the AG project produced requirement and architecture design documents for CCM3 and the coupler, a toolkit forming the basis of a next generation coupler, and a set of performance-enhanced CCM3 dynamical cores, better able to exploit microprocessor-based parallel computers.

Scope of this proposal

We propose to go well beyond the pilot project, by including all of the components of the CCSM and broadening the scope to include model development. The software design effort is expanded to the ocean, sea ice, and land-surface models. The software design must also accommodate the requirements of model development in chemistry and biogeochemistry that are included in this proposal and expected to emerge from the CCSM working groups in the next few years. Requirement and architecture design documents and implementation plans will be developed, in conjunction with the CCSM Software Engineering Working Group, to bring all components into conformity with a comprehensive software design including a machine-specific layer, a utility and library layer, and a model-code layer. This software hierarchy will enable rapid adaptation to new architectures while maintaining high production efficiencies on the available computing platforms. A modular structure will make the substitution of alternative components significantly easier than in current models.

The proposal also broadens the scope of activities to include development of new or improved model formulations, numerical algorithms, and parameterizations of physical and chemical processes; extension of component models to higher resolutions; and diagnosis, evaluation and optimization of model performance, both scientific and computational. The work under this proposal will build upon and complement model development activities under the core DOE Climate Change Prediction Program. In addition to the ongoing design work and the need to incorporate new modules and packages into the codes, we will investigate the programming paradigms best suited to exploitation of emerging architectural features of high performance computers. Coding structures and style will be adapted to ensure that the CCSM will remain a state-of-the-science model that can meet the varied demands of the climate model user community.

The scope of this proposal intentionally omits scientific applications of the model, except to the extent that scientifically interesting results are generated and extracted in the process of model evaluation and validation. Applications of the models to problems such as interannual variability, global warming, or paleoclimate are not part of the mandate of the call for proposals. They are to be carried out under other auspices. This program is intended to provide the enabling technology, in the form of model software and software that supports the model. Nevertheless, experience has shown that many interesting results do emerge during comparisons of the coupled model and its component models with the real world. Such comparisons are a critical aspect of the proposed work.

Objectives

The overarching goal is to collaboratively develop CCSM so that it is:

  • comprehensive in its treatment of physical and chemical processes important to the climate system;
  • comprised of modular 'packages' with well-defined interfaces that can be tested off-line and interchanged with packages containing updated or alternate treatments of the same processes, and
  • performance optimized yet portable and adaptable for new computing architectures.

Standardizing the model's interfaces, to simplify upgrading of component models in the coupled model and of dynamical cores and physical parameterizations in the component models, is important for a community model. The final product of this proposal will be a high quality, comprehensive model of the Earth's climate system suitable for research and as a tool for integrated assessments. We will utilize and build on the advanced dynamical cores, flexible model coupling method, and extensive toolkit developed for the AG pilot project.