Geoscience Reference
In-Depth Information
1200 polar bears in the Northern Beaufort Sea subpopula-
tion [ Aars et al. , 2006], but numbers of bears in the rest of
this ecoregion are poorly known. There are no estimates for
the east Greenland subpopulation, but we assumed there
currently may be up to 1000 bears there. We modified the
PBSG recognized subpopulation boundaries of this ecore-
gion by redefining a Queen Elizabeth Islands subpopulation
(QE). QE had formerly included the continental shelf region
and interisland channels between Prince Patrick Island and
the northeast corner of Ellesmere Island [ Aars et al. , 2006].
We extended its boundary to northwest Greenland. This area
is characterized by heavy multiyear ice, except for a recur-
ring lead system that runs along the Queen Elizabeth Islands
from the northeastern Beaufort Sea to northern Greenland
[ Stirling , 1980]. Over 200 polar bears could be resident here,
and some bears from other regions have been recorded mov-
ing through the area [ Durner and Amstrup , 1995; Lunn et
al. , 1995]. like the Northern Beaufort Sea subpopulation,
QE occurs in a region of the polar basin that recruits ice as it
is advected from the PBDE [ Comiso , 2002; Rigor and Wal-
lace , 2004; Belchansky et al. , 2005; Holland et al. , 2006;
Durner et al. , 2008; Ogi and Wallace , 2007; Serreze et al. ,
2007]. Assuming these rough estimates are close, up to 2400
bears might presently occupy the PBCE.
We did not incorporate the central Arctic Basin into our
analyses. This area was defined to contain a separate sub-
population by the PBSG in 2001 [ Lunn et al. , 2002] to
recognize bears that may reside outside the territorial ju-
risdictions of the polar nations. The Arctic Basin region is
characterized by very deep water which is known to be un-
productive [ Pomeroy , 1997]. Available data are conclusive
that polar bears prefer sea ice over shallow water (<300 m
deep) [ Amstrup et al. , 2000, 2004; Durner et al. , 2008], and
it is thought that this preference reflects increased hunting
opportunities over more productive waters. Tracking stud-
ies indicate that few if any bears are year-round residents of
the central Arctic Basin. For all of these reasons, we did not
include the Arctic Basin in our analyses.
These included hindcast ice estimates from the 20th Century
Experiment (20C3M) and projection estimates for the 21st
century forced with the “business as usual” Special Report
on Emissions Scenarios (SRES) A1B emissions scenario
[ Nakićenović et al. , 2000]. We obtained GCM ice projection
outputs of nine models from the World Climate Research
Programme's Coupled Model Intercomparison Project phase
3 (CMIP3) multimodel data set [ Meehl et al ., 2007a]. We ob-
tained projections from the 10th model (Community Climate
System Model, version 3 (CCSM3)) directly from the Na-
tional Center for Atmospheric Research in its native CCSM
grid format (D. Bailey and M. holland, NCAR, personal
communication, 2007). We obtained and analyzed one run
(run 1) for each GCM, except CCSM3 for which we obtained
eight runs. In our analyses we included the mean of the eight
CCSM3 runs as a single member of our 10-model ensemble.
We selected the 10 GCMs from a larger group of 20 based
on their ability to simulate (20C3M) the mean Northern
hemisphere ice extent for September 1953-1995 to within
20% of the observed September mean (had1SST [ Rayner
et al. , 2003]). This selection method emulated that used by
Stroeve et al. [2007], except we used a 50% ice concentra-
tion threshold [ DeWeave r, 2007] to define ice extent (as op-
posed to 15%). We chose a 50% threshold because other
studies have shown that polar bears prefer medium to high
sea ice concentrations [ Arthur et al. , 1996; Ferguson et al. ,
2000; Durner et al. , 2006, 2008].
Sea ice grids among the 10 GCMs we analyzed had vari-
ous model-specific spatial resolutions ranging from ~1 ´ 1 to
3 ´ 4 degrees of latitude ´ longitude. To facilitate integration
with our analyses of observational data, we resampled the
GCM grids to match the gridded 25 km resolution passive
microwave sea ice concentration maps from the National
Snow and Ice Data Center. Each native GCM grid of sea
ice concentration was converted to an Arc/Info (version 9.2;
ESRI, Redlands, California, United States) point coverage
and projected to polar stereographic coordinates (central
meridian 45°W, true scale 70°N). A triangular irregular net-
work (TIN) (Arc/Info) was created from the point coverage
using ice concentration as the z value, and a 25 km pixel
resolution grid was generated by sampling the TIN surface.
Effectively, this procedure oversampled the original GCM
resolution using linear interpolation.
2.3. Sea Ice Habitat Variables
Our BN model incorporated changes in area and spa-
tiotemporal distribution of sea ice habitat along with other
“stressors” that might help predict the future of polar bears.
We used monthly averaged ice concentration estimates de-
rived from passive microwave satellite imagery for the ob-
servational period 1979-2006 [ Cavalieri et al. , 1996]. Sea
ice data for the future were derived from monthly sea ice
concentration projections of 10 GCMs. The GCMs we used
were included in the Intergovernmental Panel of Climate
Change (IPCC) Fourth Assessment Report (AR4) (Table 1).
2.3.1. Total annual habitat area. For input to our models,
we defined two area-based metrics of habitat availability to
polar bears. The first was an expression of the yearly extent
of “total available ice habitat,” and the second, which was
available in the polar basin only, was an expression of “to-
tal optimal habitat.” We derived “total available ice habitat”
from both observed and projected Arctic-wide sea ice con-
Search WWH ::




Custom Search