Geoscience Reference
In-Depth Information
use of AGCM to study climate anomalies through numerical experimentation
with various prescribed forcing functions in the atmosphere, land, and oceans
(Manabe and Wetherald
1975
; Manabe et al.
1979
; Gilchrist
1977
,
1981
).
A climate model differs from a weather prediction model in that the former
has to be integrated for an extended period of time (multi-years), whereas the
latter is generally integrated for a few days at the most. Some of the current
climate models to study global change have carried out integration up to
thousands of simulated years in order to determine the reliability of long-term
climate signals. Because of the requirement for long-term integration, climate
models are most sensitive to the conservation of mass, energy, and moisture.
Small imbalance in any of the conserved properties can introduce substantial
errors that may amplify during the course of the integration to produce severe
model systematic bias - a problem known as ''climate drift.'' In contrast, for
numerical weather prediction, the accuracy of the initial conditions is more
critical, and simulations generally cover a period too short for the climate
drift to be an issue.
One of the problems facing the climate modeling community in the 1970s
and 1980s was the enormous demand on computation resources required to
carry out long-term climate simulations. As a result, for most early applica-
tions, climate models with coarse resolutions of the order of 250-500 km,
with 2-10 vertical layers, were used and the simulation periods limited to a
few years. At such coarse resolution, many physical processes are grossly
under-represented. For this reason, many of the early climate model results
can only be regarded at best as exploratory. With the advent of computer
technology, and more efficient computation codes, climate models can now
be run at increasing spatial and temporal resolutions, and with ever more
complex physics modules. At present, climate models are currently being
run with resolution higher than 50 km at operational centers such as
ECMWF. Integrations have been carried out for hundreds of years, such
as those used for the IPCC climate assessments, and other global change
scenarios by many climate modeling groups around the world. Currently,
the Earth Simulator Project at the Frontier Global Change Research Program
of Japan is running global climate models at approximately 10 km resolu-
tion, with over 200 layers in the vertical and for hundreds of simulated
years.
However, even with very high-resolution climate models, large uncertain-
ties remain with regard to prediction of future climate change, especially in
the projection of statistics of increased hazards on regional and subregional
scales due to extreme weather events. This shortcoming stems from our very
limited understanding of the physics of the real climate system, which makes
it impossible to include all the details required at the higher model resolution.
Hence, merely increasing resolution is not a panacea to the problems of
Search WWH ::
Custom Search