A Bayesian Network Modeling Approach to Forecasting the 21st Century Worldwide Status of Polar Bears - Arctic Sea Ice Decline: Observations, Projections, Mechanisms, and Implications - page 230

PBDE, by comparison, some bears retreat to shore, while
most follow the sea ice as it retreats far offshore in sum-
mer. The fact that ecoregion and the availability of alternate
ecoregions together explained 22% of the variation in overall
population outcome was further evidence of the importance
of sea ice habitat and its regional differences.
Another habitat variable, “foraging habitat character”
(node S1), was ranked 7th among variables having influence
on the overall population outcome. This qualitative variable
relating to sea ice character was included to allow for the
fact that in addition to changes in quantity and distribution
of sea ice, subtle changes in the composition of sea ice could
affect polar bears. For example, longer open water periods
and warmer winters have resulted in thinner ice in the po-
lar basin region [ Lindsay and Zhang , 2005; Holland et al. ,
2006; Belchansky et al. , 2008]. Fischbach et al. [2007] con-
cluded that increased prevalence of thinner and less stable
ice in autumn has resulted in reduced sea ice denning among
polar bears of the southern Beaufort Sea.
Observations during polar bear field work suggest that the
thinning of the sea ice also has resulted in increased rough-
ness and rafting among ice floes. Compared to the thicker
ice that dominated the polar basin decades ago, thinner ice
is more easily deformed, even late in the winter. Whether or
not thinner ice is satisfactory for seals, the extensive areas of
jagged pressure ridges that can result when ice is more eas-
ily deformed may not be well suited to polar bear foraging.
These changes appear to reduce foraging effectiveness of
polar bears, and it is suspected the changes in ice conditions
may have contributed to recent cannibalism and other unu-
sual foraging behaviors [ Stirling et al. , 2008]. Also, thinner,
rougher ice, interspersed with more open water, may be an
impediment to the travels of young cubs. Physical difficul-
ties in navigating this “new” ice environment could explain
recent observed increases in mortality of first-year cubs
[ Rode et al. , 2007]. The fact that six of the seven variables
most influential on overall outcome were sea ice related and
explained 87% of the variation in that outcome corroborates
the well established link between polar bears and sea ice.
The 5th ranked potential stressor to which overall popu-
lation outcome was sensitive was intentional takes. his-
torically, the direct killing of polar bears by humans for
subsistence or sport has been the biggest challenge to polar
bear welfare [ Amstrup , 2003]. Our model suggests that har-
vest of polar bears may remain an important factor in the
population dynamics of polar bears in the AE and PBCE, as
sea ice retreats.
It is important to remember that there is great uncertainty
in the exact way the potential stressors we modeled may
change in the future. Also, the degree of uncertainty differs
among the variables we included in our model. There is rela-

tively great certainty that the spatiotemporal distribution of
sea ice will decline through the coming century. There is
less certainty in just how much it will decline by a speci-
fied decade. The short, intermediate, and long-term effects
of that decline on food availability for polar bears are largely
unknown [ Bluhm and Gradinger , 2008]. here, we assumed
that declining sea ice means declining food availability for
polar bears, with the decline in food mirroring that of the
sea ice. Spatiotemporal reductions in sea ice cover, how-
ever, could fundamentally alter the structure and function
of the Arctic ecosystem. Such changes could result in dif-
ferent timing and level of productivity. It seems clear that
continued declines in sea ice ultimately will mean reduced
year-round food availability and declines in polar bear num-
bers and distribution. We cannot rule out, however, that in-
creases in productivity could result in transitory increases in
food availability for bears. Such changes would not alter the
ultimate predictions made here, polar bears are clearly tied
to the sea ice for access to their food. Such changes could,
however, alter the temporal sensitivity of our outcomes to
values at input nodes.

4.4. Strength of Evidence

Our BN population stressor model projects that sea ice
and sea ice related factors will be the dominant driving force
affecting future distributions and numbers of polar bears
through the 21st century. Despite caveats regarding the early
stage of development of our BN model, there are several
reasons to believe that the directions and general magnitudes
of its outcomes are reasonable.
First, they are consistent with hypothesized effects of glo-
bal warming on polar bears [ Derocher et al. , 2004] and with
recent observations of how decreasing spatiotemporal distri-
bution of sea ice has affected polar bears in some portions of
their range [ Stirling and Derocher , 1993; Stirling and Par-
kinson , 2006; Stirling et al. , 1999, 2007, 2008; Ainley et al. ,
2003; Ferguson et al. , 2005; Amstrup et al. , 2006; Hunter
et al. , 2007; Regehr et al. , 2007a, 2007b; Rode et al. , 2007].
The high sensitivity of our overall model outcomes to sea
ice habitat changes is consistent with the recent PBSG deci-
sion, based mainly on projected changes in sea ice, that polar
bears should be reclassified as vulnerable [ Aars et al. , 2006].
Second, results of influence runs to assess the ability of on
the ground activities to alter outcomes were not qualitatively
different, during most time periods, from previous runs for
the PBDE and SIE (Plate 6). Maintaining current conditions
(other than sea ice) in the PBDE and SIE or improving con-
ditions on the ground appeared to have some ability to re-
duce the risk of extinction being the most probable outcome
during the first couple decades of this century. This effect