Collaborative Design and Development of the Community Climate System Model for Tera-scale Computers

Robert Malone, Los Alamos National Laboratory
John Drake, Oak Ridge National Laboratory
Other DOE Laboratories: Argonne, Berkeley, Livermore, and Pacific Northwest

Summary

1. Introduction

Evidence continues to accumulate of anthropogenic influences on global climate. While extensive monitoring and thorough analysis of observations are essential, it is also crucial that “fully coupled” climate simulation models be brought to the highest level of fidelity. Physical, chemical, and biological processes operating in the four component models (atmosphere, ocean, sea ice, and land surface) must be accurately represented. Much finer spatial resolution is needed to obtain meaningful projections of regional climate change and socioeconomic impacts. Finally, the chaotic nature of the atmosphere and ocean means that ensembles of runs with perturbed initial conditions are needed to quantify the range of uncertainty.

The NCAR Community Climate System Model (CCSM) is a state-of-the-art climate model. Its user community is arguably the largest of any climate model in the world. Although CCSM development was primarily funded by the National Science Foundation, DOE has contributed to the NCAR climate modeling effort since the early 1980's, when the first Community Climate Model was developed. DOE became more extensively involved under the CHAMMP program, which helped bring massively parallel computing to climate modeling. DOE's contributions were refocused under the Climate Change Prediction Program (CCPP) and broadened in this SciDAC project.

2. Major Objectives

The major objectives of the Consortium are: (A) improving the parallel performance of the CCSM and its component models, while (B) maintaining their portability on a wide range of computer architectures; (C) upgrading the software engineering practices applied to the component models and the coupler; (D) developing a new generation of ocean model based on a hybrid vertical coordinate that resolves the turbulent surface mixed layer but also follows isopycnal surface in the density-stratified deep ocean; (E) adding atmospheric chemistry; and (F) adding ocean biogeochemistry. The last two items are necessary for accurately modeling the carbon cycle, carbon sequestration, aerosol effects and alternative emission and mitigation scenarios.

3. Recent Progress

This past year has been both exciting and challenging for our Consortium members. Through a collaboration with Japanese scientists, large blocks of time on the Earth Simulator system are available for CCSM runs targeted for the 2007 IPCC (Intergovernmental Panel on Climate Change) climate assessment. The Consortium hosted two vectorization workshops and worked closely with NCAR and Japanese scientists as well as vendor representatives to (re)introduce vectorization into the CCSM and its component models. This has been done, so the CCSM3 is now ready to carry out the important simulations needed by the IPCC assessment of global warming.

Steady progress has been made on the new hybrid-coordinate ocean model (HYPOP). A new formulation of the vertically averaged “barotropic” mode has led to a two-time-level subcycled explicit solution, in place of the fully implicit scheme used in the Parallel Ocean Program (POP), developed at Los Alamos as the ocean component of CCSM. A geodesic-grid version of HYPOP will be the ocean component of a geodesic climate model being developed by David Randall (CSU) under SciDAC sponsorship.

NCAR, LLNL, ANL and ORNL have been working closely on implementing a compact chemistry scheme in the atmospheric model. This team is working smoothly and making good progress. Ocean biogeochemistry is an even more complicated system, but significant progress has been made in implementing a more community-based biogeochemistry model in collaboration with the Paradigm project funded by the National Ocean Partnership Program. Advances published by SciDAC subteams will be incorporated as improvements to this model.

4. Peer Review

A peer review of this SciDAC Consortium was held at Oak Ridge in October 2003. The review was quite favorable, as in this quotation from their report: “SciDAC has effectively brought together teams of research scientists and computational experts from DOE national laboratories and NCAR, and has provided a single major focus for their research. In the process, this capability is becoming a national asset.”

It is, in fact, the diversity of talent that has been assembled in this project that makes it so effective. The combination of software engineers, applied mathematicians, together with a wide range of physical scientists has enabled the Consortium to contribute to all aspects of the CCSM. We have benefited greatly from our interactions particularly with the PERC ISIC, and expect increasing interactions with the ESG and SDM ISICs. Ongoing projects using the CCA technology and cooperation with NASA's ESMF will allow integration of new software from the APDEC, TOPS and TSTT centers.

5. Future Plans

During the final two years of the SciDAC Consortium project, we plan to integrate the vectorization capability of CCSM more carefully and to enhance its performance-portability. We expect to have atmospheric chemistry and ocean biogeochemistry fully integrated into the model. The monotone, low diffusion “incremental remapping scheme” in HYPOP will make it well suited to transport the large numbers of tracers needed for ocean biogeochemistry. Significant steps will be taken to equip the land model with a full carbon-cycle simulation capability and to support regionally resolved climate change studies.

back to project page