Geoscience Reference
In-Depth Information
[ Persson et al. , 2002] in Figure 2. The error bars shown in
Figure 2 indicate the standard deviation over simulation
years 250-599. Although the simulated net shortwave radia-
tive flux exceeds observations in July and August by 15 to
18 W m -2 (Figure 2a), this is a significantly smaller over-
estimate than for other climate models examined by Goro-
detskaya et al. [2008] for the SHeBA location (not shown).
Additionally, the simulated surface albedo is within the
range of observations used in SHeBA. The magnitude of the
simulated net longwave flux shown in Figure 2b is gener-
ally greater than observation. Simulated sensible and latent
heat fluxes are small components of the total energy budget
in agreement with observation. However, the pronounced
annual cycle in the latent heat flux is not well reproduced
(Figures 2d and 2e).
Finally, Tremblay et al. [2007] examined simulated Arc-
tic Ocean temperature and salinity profiles in comparison to
four other climate models. The ccSm3 was found to have a
good representation of surface stratification. This is impor-
tant for simulating the effects of Atlantic warm water intru-
sions into the Arctic, which are typically insulated from the
surface ice pack by a sharply defined halocline. However,
the cold halocline layer in the model is absent (as in every
other model evaluated), and the Atlantic waters were found
to be 1°c too warm and somewhat saltier.
In summary, Arctic sea ice characteristics and 20th century
trends are well represented in the ccSm3; however, this is
achieved with anomalous cyclonic circulation in the atmo-
sphere and a larger poleward Atlantic oceanic heat transport
than observed. These discrepancies highlight the difficulties
in simulating the Arctic climate system.
running means of winter and summer sea ice extent, suggest-
ing the presence of low-frequency climate variability that af-
fects the ice cover in both seasons. The correlation between
the two filtered time series is 0.28 over the full period. If
a period of low march ice cover during the years 350-400 is
excluded, however, this correlation is then 0.61. This period
of low march ice cover (year 350-400) is associated with
anomalous concentrations in the Sea of Okhotsk.
From visual inspection of Figure 3, it may be seen that there
are extended periods of both positive and negative trends; it
is noted that there are extended periods of the simulation
where the magnitude and trend of the September minimum
operates within the bounds of 1979-2006 observed sea ice
extent. For example, September ice extent decreases for the
years 300 to 340 at a rate of 6% per decade. This is similar
to the observed trend of 8% per decade over the 1979-2006
satellite era [ Stroeve et al. , 2007]. Of particular interest are
three marked events occurring in model years 451, 490, and
556-558. each of these events is approximately 3 standard
deviations less than the simulated long-term September
mean extent, and they roughly match the observed dramatic
minimum which occurred in September 2007, with the first
two events being slightly less than this value. These events
largely occur in the absence of significant ice anomalies
during the preceding winter months. For example, the third
event occurs over three consecutive September months in
which ice anomalies are near zero or are positive in the in-
tervening months of November through April of each year,
as referenced to the 350-year simulation period examined.
The monthly anomalies for each event occur in July through
October, with September incurring the largest anomaly. The
spatial distribution of these three events is shown in Plate 3.
For reference, the simulated long-term average September
extent is shown as a red line.
From Plate 3, it may be seen that the ice loss for the first
two events occurs primarily on the Pacific side of the basin.
For model year 451, the losses are particularly evident in the
chukchi, laptev, and east Siberian seas. For model year 490,
September ice cover is notably absent in the Beaufort, chuk-
chi, east Siberian, and laptev seas, while the third event
shows losses primarily adjacent to the Siberian coastline and
in the central Arctic. The spatial distribution of these losses
are reminiscent of dramatic reductions in the summertime
ice cover from the recent observational record of 2002-2007
[ Ogi and Wallace , 2007]. These simulated events are now
examined in more detail.
3. reSulTS
Figure 3 (top) shows the time series of the September
Northern Hemisphere sea ice extent from the CCSM3 control
simulation for model years 250 through 599 using constant
1990 solar, trace gas, and aerosol forcing. For reference, the
record maximum and minimum values of September sea ice
extent from the HadISST1 climatology [ Rayner et al. , 2003]
from the period 1979-2006 are shown along with the record
September 2007 minimum extent. There is considerable vari-
ability in the time series, and occasionally the 2007 minimum
extent value is reached. In Figure 3 (bottom), it may be seen
that the interannual variability in September sea ice extent is
larger than for March, the month of maximum ice extent. The
standard deviation is 0.32 × 10 6 km 2 for the March time se-
ries and 0.54 × 10 6 km 2 for the September time series. This is
in agreement with the available satellite record, which shows
greater variability and trends in the September sea ice extent
[ Serreze et al. , 2007]. There is a correlation between 20-year
3.1. Simulation Year 451 Event
As described previously, this event is characterized by
substantial ice cover losses throughout the Arctic Basin but
Search WWH ::




Custom Search