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 233

largely disappears by mid century, however, and manage-
ment of localized human activities appears to have little
qualitative effect on the future of polar bears in the PBDE
and SIE if sea ice continues to decline as projected. The fact
that sea ice has been declining more rapidly than even our
minimum ice GCMs project (see below) suggests even pos-
sible transitory benefits of on the ground interventions may
be illusory in these ecoregions. In contrast, our BN model
suggested that managing human controlled stressors could
qualitatively lower the probabilities of extinction in the
PBCE and AE. Such management has played an important
role in the past recovery of polar bears [ Amstrup , 2003] and
apparently could continue to be important in regions where
sea ice habitat remains.
Third, influence runs in which we specified complete un-
certainty in all inputs except sea ice illustrated that in order
to obtain any qualitative change in the probability of extinc-
tion in any of the ecoregions, sea ice availability in future
time periods must be greater than even the maximum-ice
GCM projections we used (Plate 6). This eventuality may be
unlikely in light of the fact that most GCMs simulate more
ice than has actually been observed during 1979 to present
[ Durner et al. , 2008; Stroeve et al. , 2007]. It also seems un-
likely in light of the most recent observations that September
sea ice extent (15% concentration) in 2007 was 4.13 ´ 10 6
km 2 . This is lower than the previous record set in 2005 by
nearly 1.2 ´ 10 6 km 2 (National Snow and Ice Data Center,
1 October 2007 press release, available at http://nsidc.org/
news/index.html). When ice extent based on a 50% concen-
tration threshold, which is probably near the lower limit of
ice cover useful to polar bears [ Durner et al. , 2008], is calcu-
lated, the area in 2007 was down to approximately 3.5 ´ 10 6
km 2 . Five of the GCM models in our 10-member ensemble
forecast more September sea ice in 2050 than was observed
in 2007 (Plate 8). Perhaps even more telling than the overall
“faster-than-forecasted” ice decline, is the observation that
much of the AE, the region where we forecasted polar bears
to remain until late in the century, was ice free in September
2007 (Plate 8). hence, the probability that more sea ice than
projected will be available during the rest of this century
seems low.
Finally, a polar bear BN population stressor model would
have to be structured and parameterized very differently to
project qualitatively different outcomes than we have here.
yet it seems unlikely that other polar bear experts would
do that. Evidence for the polar bear's reliance on sea ice is
replete. Although they are opportunistic and will take ter-
restrial foods, including human refuse when available, and
may benefit from such activity [ Lunn and Stirling , 1985; De-
rocher et al. , 1993], polar bears are largely dependent on the

productivity of the marine environment. Refuse, for exam-
ple, is of limited availability throughout the polar bear range,
and could at best benefit relatively few individuals. Also,
polar bears are inefficient in preying on terrestrial animals
[ Brook and Richardson , 2002; Stempniewicz , 2006]. Per-
haps most importantly, polar bears have evolved a strategy
designed to take advantage of the high fat content of marine
mammals [ Best , 1984]. Available terrestrial foods are, with
few exceptions, not rich enough or cannot be gathered ef-
ficiently enough to support large bodied bears [ Welch et al. ,
1997; Rode et al. , 2001; Robbins et al. , 2004]. Because polar
bears are the largest of the bears, it is unlikely that terrestrial
arctic habitats which are depauperate from the standpoint
of bear food could support them in anything like current
numbers. Empirical evidence of this is provided by the fact
that Arctic grizzly bears are the smallest grizzly bears found
anywhere and they occur at the lowest densities [ Miller et
al. , 1997]. habitats adjacent to present polar bear ranges just
do not seem likely to support large numbers of the much
larger-bodied polar bears. Although polar bears in Hudson
Bay, which are forced onto land all summer, are known to
consume a wide variety of foods; they gain little energetic
benefit from those foods [ Ramsay and Hobson , 1991]. The
only foods, other than their ringed and bearded seal staples,
that are known to be energetically important to polar bears
in some regions are other marine mammals [ Iverson et al. ,
2006]. Polar bears, it appears, are obligately dependent on
the surface of the sea ice for capture of the prey necessary to
maintain their populations.
In short, although other polar bear experts might structure
a model differently and populate the conditional probability
tables differently than we have here, it seems unlikely that
those differences would be great enough to make a qualita-
tive difference in the outcomes projected by our BN popula-
tion stressor model for mid century and beyond.

5. CONClUSIONS

We used a first-generation BN population stressor model
to forecast future populations of polar bears worldwide. Out-
comes of this model suggested that declines in the spatiotem-
poral distribution of sea ice habitat along with other potential
stressors will severely impact future polar bear populations.
Polar bears in the PBDE and SIE, home to approximately
two thirds of the current world population will likely dis-
appear by mid century. Management of localized human
activities is unlikely to mitigate those mid century losses.
Polar bears in and around the AE appear likely to persist
late into the 21st century, especially if on-the-ground human
activities, particularly human harvests, are carefully managed.