Geoscience Reference
In-Depth Information
Figure 9.13. Multimodel ensemble mean sea
ice extent based on observations (black line),
the CMIP5 models (red line for the ensemble
mean and red shading for the +/− 1 standard
deviation) and the CMIP3 models (blue line for
the ensemble mean and red shading for the +/−
1 standard deviation) (courtesy of J. Stroeve,
NSIDC, Boulder, CO). (See plate section for
color version.)
before the end of the twenty-first century. Viewed in this context, conclusions
drawn from CMIP5 are not much different than those from CMIP3. Importantly,
the CMIP5 simulations do not appear to have appreciably reduced uncertainty as
to when a seasonally ice-free Arctic Ocean will be realized. In a subsequent study,
J. Liu et al. ( 2013 ) attempted to reduce the spread in the projecting timing of a sea-
sonally ice-free Arctic Ocean (ice extent less than 1.0 × 10 6 km 2 ) based on both the
ability of models to reproduce the observed sea ice climatology and variability since
1979 and the strong and persistent relationships between present and future ice
conditions. Both approaches point to a seasonally ice-free state between 2054-2058
assuming a high emissions (RCP 8.5) scenario.
Understanding why different models behave differently, even when driven with the
same climate forcings, is a very active area of research. With respect to the tendency
for the CMIP3 models to be “conservative” regarding their depiction of the trend in
September extent over the period of observations, it appears that at least part of the
explanation is that these models do not capture the observed thinning of the ice cover
and the acceleration in drift speed that should accompany this thinning (Rampal et al.,
2011 ). With respect to the wide range in projections of sea ice extent, much attention
has focused on depictions of mean sea ice thickness and its spatial distribution under
the present climate. Differences in mean thickness between different models point,
at least in part, to difference in the sea ice energy balance, in turn pointing to issues
such as cloud cover as it affects the downward solar and longwave radiation fluxes, or
ocean heat fluxes. Differences in the special distribution of ice thickness point to dif-
ferences in depiction of the surface wind field. Simply put, if a model cannot simulate
the mean sea level pressure field over the Arctic Ocean with reasonable fidelity, the
spatial pattern of sea ice thickness is likely to be poorly simulated.
Along with considerable scatter in depictions of present-day ice thickness and
extent, results from the CMIP5 simulations point to some systematic model biases.
Figure 9.14 shows the spatial pattern of observed September sea ice thickness based
on the NASA ICESat aircraft missions (aircraft altimetry), the September pattern
based on averaging output from all of the CMIP5 models for years corresponding
to the ICESat data, and the difference between the modeled and observed thickness.
The ICESat data depict the well-known pattern of the thickest ice being located
north of the Canadian Arctic Archipelago and Greenland, with thinner ice over the
Eurasian side of the Arctic Ocean. The difference field (modeled minus observed
thickness) shows clearly that, viewed as a group, the ice cover from the CMIP5
Search WWH ::




Custom Search