|
Michael Wehner, Lawrence Livermore
National Laboratory
Research Objectives
The principal mission
of the Program for Climate Model Diagnosis and Intercomparison (PCMDI)
is to develop improved methods and tools for the diagnosis, validation,
and intercomparison of global climate models (GCMs) and to engage in research
on a wide variety of outstanding problems in climate modeling and analysis.
In this project, we are using the T3E to expand a set of ensemble climate
integrations of the coupled ocean-atmosphere model of the Hadley Centre
in the United Kingdom. Specifically, we are investigating the linearity
of the climate's response to different kinds of forcing. Previous integrations
performed in Great Britain are being augmented to include integrations
with only anthropogenic sulfate aerosol forcing.
Computational Approach
Much of the previous
year was spent porting the model (HADCM2) from the UK Meteorological Office
T3E to the NERSC T3E. The model is extremely large, both in number of
Fortran lines and K-shell scripting lines. The porting is complete and
the model is currently running. The model is not highly parallel, running
at only 24 processors. Model runs are extremely lengthy, requiring about
32,000 node hours per integration. However, additional parallelism is
obtained by running more than a single integration at a time.
Accomplishments
About 50 years of
integration have been performed, which is about one-eighth of the requirement
for this project.
Significance
The coupled climate system
is a complex nonlinear system. Our understanding of the climate response
to different types of forcing, e.g., enhanced carbon dioxide concentrations,
variations in solar forcing, sulfate aerosols, etc., comes largely from
integrations of global coupled GCMs. The chaotic nature of the climate system
requires us to perform ensembles of integrations to separate response signal
from climatic noise. This work will complete the missing parts of the most
comprehensive set of coupled model ensembles.
Publications
http://www-pcmdi.llnl.gov/
 |
|
Number of simulation runs required to determine decadal mean December
through February surface air temperature within 1.0°K at 95% statistical
confidence. Results are shown for temperatures obtained from the NCAR
Parallel Coupled Model. Ensembles of climate simulations enable researchers
to quantify climate variability and better predict the uncertainty
of future climate change. Unfortunately, the high computational costs
of fully coupled climate models limit the size of such ensembles.
Our research has provided a means of determining how many ensemble
realizations are required in advance of performing the calculations.
This number is dependent on which climate variable is of interest
and on the desired accuracy. |
|
