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 220

Table 2. (continued)

Node D1: Overall Population Outcome

Time
Period

Most Probable
Outcome

Same as
Now (%)

Basis larger (%) Smaller (%) Rare (%) Extinct (%)

Polar Basin Convergent Ecoregion

Year −10 Satellite data larger 98.39 1.61 0.00 0.00 0.00

year 0 Satellite data larger 71.69 27.49 0.63 0.19 0.00

year 25 GCM minimum larger 64.86 28.47 6.09 0.49 0.09

year 50 GCM minimum extinct 0.26 2.30 27.98 31.59 37.87

year 75 GCM minimum extinct 0.00 0.39 9.68 13.24 76.70

year 95 GCM minimum extinct 0.00 0.39 9.68 13.24 76.70

year 25 Ensemble mean same_as_now 18.23 41.81 26.37 5.16 8.43

year 50 Ensemble mean extinct 0.48 2.72 29.27 32.46 35.06

year 75 Ensemble mean extinct 0.00 0.27 8.40 15.10 76.23

year 95 Ensemble mean extinct 0.02 0.44 9.49 12.75 77.30

year 25 GCM maximum same_as_now 18.23 41.81 26.37 5.16 8.43

year 50 GCM maximum extinct 0.14 1.24 21.15 30.71 46.77

year 75 GCM maximum extinct 0.02 0.46 12.64 24.46 62.41

year 95 GCM maximum extinct 0.02 0.44 10.51 16.52 72.52

a See Plate 3.

the earliest time periods, we forecasted low probabilities for
all other outcome states in these two ecoregions. The low
probability afforded to outcome states other than extinct
suggested a clear trend in these ecoregions toward prob-
able extirpation by mid century. Forecasts were less severe
in other ecoregions. At year 50, probability of the “extinct”
state was only 8-10% in the AE. At all time steps in the
AE, and at year 50 in the PBCE, considerable probability
fell into outcome states other than extinct (Table 2 and Plate
4). The distribution of probabilities for the states of overall
population outcome suggests polar bears could persist in all
ecoregions through the early part of the century, through mid
century in the PBCE and through the end of the century in
the AE (Table 2 and Plate 4).
Future conditions affected node C3, polar bear distribution,
more than they affected node C4, polar bear numbers. “Ex-
tirpated” was the most probable outcome at mid century for
node C3 in the PBDE and SIE. The most probable outcome
in these ecoregions for node C4, however, was reduced den-
sity [see Amstrup et al. , 2007]. This probably reflects the high
relative certainty that areas where ice is absent for too long
will not support many bears, and the relative uncertainty re-
garding how population dynamics features may change while
the sea ice is retreating. Modeled future polar bear distribu-
tions were driven by the FWS listing factor “Habitat Threats”
(node F2, Table 3), as well as by node D, habitat distribu-
tion. Distribution and availability of habitat, especially in the

SIE and PBDE (Table 3) appear to be the most salient threats
to polar bears. We also assumed that deteriorating sea ice
would be accompanied by worsening conditions listed un-
der FWS listing factors C, disease and predation, etc., and E,
other natural or man-made factors (nodes A4 and A6) (Table
4). We included year 25 in our projections to help provide
context for mid century projections and beyond and to help
understand the transition from current to future conditions. It
is important to emphasize, however, that polar bears have a
long life span. Many individuals alive now could still be alive
during the decade of 2020-2029. hence, projecting changes
between now and then incorporates the uncertainty of trade-
offs between functional and numerical responses, as well as
the greater uncertainties in sea ice status in the nearer term.

3.2. Sensitivity Structure of the Bayesian Network Model

We conducted 10 tests on the BN population stressor
model to determine its sensitivity structure (Appendix C).
The BN model was well balanced in that sensitivity of over-
all population outcome (node D1, sensitivity test 1) was not
dominated by a single or small group of input variables.
Considering that “ecoregion” and “availability of alternate
regions” are in essence habitat variables, 6 of the top 7 vari-
ables explaining overall outcome were sea ice related and
together explained 87% of the variation in overall popula-
tion outcome (node D1, Appendix C and Plate 5).