Michael Wehner, Justin J. Hnilo, James Boyle, Benjamin D. Santer, Karl Taylor, and Balasubramanian Govindasamy, Lawrence Livermore National Laboratory

Time series of the differences in global temperature anomalies from an uncoupled to a coupled mode (AMIP-CSM). Global values between -0.1 and +0.1 are not colored to allow for ease of comparison; values are derived over 491 months.

Research Objectives
In the Atmospheric Model Intercomparison Project (AMIP), an atmospheric general circulation model (AGCM) is forced with observed sea surface temperatures (SSTs) and sea ice in the hopes that the model will mimic observed atmospheric behavior. Results from participant modeling groups in AMIP I and II show that these prescribed boundary conditions appear to be sufficient in supplying enough information to an AGCM to allow it to reasonably represent recent atmospheric behavior. Some concerns have been raised about the use of prescribing SSTs in a model integration. For example, the specified SSTs might alter the spatial or temporal variability of the simulation compared to a coupled run and presumably reality. This effect is expected to be most prominent in the mid to high latitudes.

This research directly addresses this issue by using the SSTs from the Coupled System Model (CSM) run, which is a fully coupled Ocean-Atmosphere General Circulation Model (OAGCM), as boundary conditions to force the identical atmospheric component alone. Quantifiable differences between the two runs will be attributable in part to the use of prescribing SSTs (as a representation of the ocean) versus using a fully coupled (OAGCM) run. A simple yet important question that we can quantify and answer is "Does the ocean component in this model act to diminish or enhance atmospheric variability when compared to a prescribed SST run and observations?" This research will highlight variables and regions of greatest magnitude and variability changes.

Computational Approach
This model is a spectral transform atmospheric general circulation model. One-dimensional message passing, although simple, limits its scalability to 64 processors at the resolutions we are interested in. We are using the Community Climate Model (CCM 3.6.6), the state of the art in general circulation models.

Accomplishments
We completed assembly of four 41-year, T42 resolution (128 x 64 x 17) CSM SST-driven AMIP II style integrations. Monthly data have been rewritten and archived on HPSS. Boundary condition SSTs for model continuations for the next 40+ years have been processed. A new resolution version (T239) of this atmospheric model has been ported to the IBM SP and has been shown (in some preliminary runs) to be stable. We look forward to using this new version in continuing research.

Significance
The National Center for Atmospheric Research's coupled model known as CSM is the de facto national climate model. In the Program for Climate Model Diagnosis and Intercomparison (PCMDI) at LLNL, a major thrust is being undertaken to diagnose this model. Currently, only the atmospheric portion of the model, known as CCM3, runs on moderately parallel computing systems. Hence, in addition to our activities in analyzing the CSM runs made at NCAR, we wish to augment these data with studies of its atmospheric component.

The studies on atmospheric variability relate directly to the variability obtained in the fully coupled model. It is vital to our understanding of climatic change to characterize a model's natural variability. This study will contribute to our knowledge of the sources of both observed and modeled variability.

Higher resolution models (T239) offer one the unique opportunity to resolve, as an example, the California central valley, allowing for unprecedented levels of large to regional scale model diagnostics. The results of these model diagnostics will help modelers understand the limitations of certain parameterizations and lead to next-generation model improvements.

Publications
M. F. Wehner, "Determination of the sampling size of AGCM ensemble simulations," Climate Dynamics 16, 321 (2000).

J. J. Hnilo, B. Govidasamy, M. Wehner, J. Boyle, K. Taylor, P. Duffy, and G. Potter, "Sensitivity of a GCM to a coupled and uncoupled mode" (submitted).