final_program_abstracts[1]

11 IMSC Session Program
Constraining climate sensitivity by natural interannual
variability in the CMIP3 ensemble
Thursday - Parallel Session 2
D. Masson and R. Knutti
Institute for Atmospheric and Climate Science, ETH Zurich, Switzerland
The climate sensitivities of the CMIP3 climate models used in the IPCC AR4 report
cover a range from about 2 to 5 K. Although much of the information contained in
observation datasets has already been used to calibrate these models, it is difficult to
find an observational constraint to get a more accurate uncertainty estimate of the true
climate sensitivity. Such a problem can partly be circumvented by using a large
ensemble of perturbed physics climate models where parameters have not been
calibrated to mimic observation. This technique estimates the parametric uncertainty
of a model but not the structural uncertainty. Because CMIP3 is the only ensemble
that samples structural uncertainty, it is important to derive confidence interval based
on it using observable constraints.
Wu and North (JGR, 2003) found a possible relationship between internal variability
and climate sensitivity in an older set of global climate models. Its application to the
present CMIP3 ensemble predicts a climate sensitivity between 2.45 and 3.95 K (95%
confidence interval) using the ERA40 observation data. The constraint on climate
sensitivity relies on the interannual variability surface temperature of individual
calendar months. The internal variability of each 12 calendar months is first computed
at the grid-point scale on a common T42 resolution and then globally averaged.
Summer months are less variable than winter months and the ratio of these two
variances correlates positively with climate sensitivity. The origin of this correlation
is still not fully clear, but understanding why internal variability correlates to climate
sensitivity may provide a new constraint to improve climate models and reduce their
spread in the future.
Abstracts 240