Abstract

1. Introduction

Polar bears (Ursus maritimus) are charismatic megafauna that symbolize the Arctic. They play an important cultural, spiritual, mystical, and traditional role in the lives of Canadian Inuit through hunting and subsequent sharing of meat and fur. Additionally, Inuit-guided sport hunts provide important revenue for the economically challenged communities (Lee and Taylor, 1994). The latest research findings suggest, however, that this multi-purpose natural resource faces threats from climatic change and environmental stress

(Stirling and Derocher, 1993; Stirling et al., 1999; World Wide Fund for Nature, 2002; Derocher et al., 2004) or from simply unsustainable harvests by human hunters (see recent discussion in Taylor et al., 2005).

Unfortunately, polar bears and their shrinking ice habitat are commonly used rhetoric to argue for the possible severity of climate change and global warming to the general public (cf., Washington Post, 2005). The polar bears that are most often cited are a specific population that inhabits the southwestern Hudson Bay coast—1 of 14 polar bear populations found in Canada (Derocher et al., 1998; Taylor et al., 2001). The area they occupy encompasses almost the southernmost extent of the species (only the southern Hudson Bay polar bear population reaches farther south; Derocher et al., 1998). Population stresses have been observed, which has led to the proposition that an earlier break-up of Hudson Bay ice (and an associated increase in spring air temperatures) is the cause of decreases in reproduction, sub adult survival, and body mass of some of these bears (Stirling and Derocher, 1993; Stirling et al., 1999). A long-term warming trend of spring atmospheric temperatures was proposed, though not shown directly,[1] to be ‘‘the ultimate factor’’ (Stirling et al., 1999, p. 294). As a result, it is commonly believed that climatic changes (or ‘‘global warming’’) are the predominant factors leading to adverse conditions for the polar bear populations, although other factors have been acknowledged (e.g., density-dependent population responses; Derocher and Stirling, 1992).

We argue that there are several related stress factors that can explain the observed patterns in polar bear population ecology. Global warming may indeed have an effect on the polar bears of western Hudson Bay (WH) but it must be assessed in a more realistic framework that considers all the likely stress factors and their cumulative impacts. In such a context, it is difficult to isolate one factor of predominant severity and, consequently, it is simply not prudent to overstate the certainty of any single factor. As emphasized in Li (2004) and Loehle (2004), a full scientific understanding of an issue as complex as the population ecology of polar bears must necessarily requires the combined assessment of both the natural and social systems rooted in the problem rather than consideration of either component in isolation (i.e., warmer spring air temperatures and related sea-ice conditions in WH).

In the next two sections, we examine some of the potential nonclimatic causes of decreased reproduction, offspring survival, and body masses, including repeated bear–human interactions, food availability and competition. We then consider climatic factors by examining available surface air temperature records and ice dynamics in the Hudson Bay basin. Finally we synthesize these findings to critically evaluate the forecasts of polar bear extinction in relation to model projected scenarios of global warming by Derocher et al. (2004).

2. Human–polar bear interactions in western Hudson Bay

Western Hudson Bay polar bears have a long history of interactions and confrontations with humans. Stirling et al. (1977) discusses interactions between humans and WH polar bears from Churchill at dump sites, in town, and adjacent town areas. Over the years, the three main sources of bear–human interactions for the WH bears are activities related to (a) scientific research, (b) tourism, and (c) the Polar Bear Alert Program.

Research activities for the WH area began in 1966, and continue today as a long-term ecological monitoring project in which over 80% of the bear population is marked (Stirling et al., 1977; Lunn et al., 2002). The majority of this field work has been carried out by the Canadian Wildlife Service (CWS), although universities also conduct research on polar bears in the area. Many bears are captured, marked, and eventually recaptured, sometimes within the same year, over a number of years (e.g., Calvert et al., 1991a, b, 1995a, b, 1998). For example, from 1977 to1995, an estimated total of 2772 bears were captured (Derocherand Stirling, 1995, their Tables 2 and 3; Lunn et al., 1997a, their Tables 2 and 3), with a minimum (i.e., since not all captures are clearly reported in publications and conflicting information exists) of about 1100 recaptures (recapture rates of between 52and 90%; mean number of bears captured/year between 1977and 1995 is about 145 bears; see summary total of columns 2and 3 in Table 1). If one considers that the WH population estimate then was between 700 and 1200 bears (Amstrup and Wiig, 1991; Wiig et al., 1995), and about 15–30% of the population was captured and recaptured due to high fidelity to locations along the coast (Derocher and Stirling, 1990a, b), it is very likely that many bears were/are exposed to capture activities on a repeated basis.

An assumption most frequently made by researchers is that their work (i.e., capturing and handling wildlife repeatedly) has no significant effect on fitness, behavior or survival of the wildlife species in question (Seber, 1973; Lehner, 1979). Long term trends of handling polar bears were suggested by Ramsayand Stirling (1986) and included the possible effects on females with cubs. Although their study did not find any statistically significant results, the trends they presented indicated that females may suffer from handling by being displaced from feeding sites, possibly resulting in lowered body mass. Note that female polar bear body mass is positively related to cub survival feeding sites, possibly resulting in lowered body mass. Note that female polar bear body mass is positively related to cub survival (Derocher and Stirling, 1996, 1998a). If females lose body mass due to handling, cubs will be adversely affected in their survival rates. Also, most polar bear capture work occurs either on family groups in spring as they emerge from their dens, or during the ice-free period while bears are distributed along the southwestern shore of either stressed due to lactation (Arnould, 1990) or undergo a fasting period while living off their stored fat reserves (Watts and Hansen, 1987). While the handling effect study of Ramsay and Stirling (1986) covered only 1967–1984, we suggest an additional analysis of capture–recapture data for handling effects that extends their time period to the present.

Almost concurrently with research activities at WH, some of the bears in the WH population are exposed to tourists and tourism activities during the fall. Since about 1980, polar bear viewing from large customized vehicles has been practiced near the town of Churchill. Polar bears leave the ice during June/July and slowly migrate north to the shores of Hudson Bay (approximately 35 km east of Churchill) where they congregate and wait the early freeze-up of the Bay, usually during November. Tour companies transport visitors into the congregation area (approximate coordinates are: 588450N to588480N, and 938380Wto 938500W) during October/November to view the bears (Dyck, 2001). Although the viewing period is short, usually between 1 October and 15 November, it is very intense, with about 6000 tourists and 15 large tundra vehicles per day in the area (Dyck and Baydack, 2006). Baiting, harassment and chasing of bears have been documented to occur (Watts and Ratson, 1989; Herrero and Herrero, 1997). ThePolar Bear Technical Committee has expressed concern over these activities, suggesting that harassment of bears during this time of the year might be very stressful due to their fasting (Calvert et al., 1998). In the first baseline study conducted in the area to address tundra vehicle behavior and vigilance (i.e., a motor act that corresponds to a head lift interrupting the ongoing activity) of resting polar bears, Dyck and Baydack(2004) found significant increases in vigilance behaviour of resting male polar bears in the presence of vehicles. The authors speculated that increased vigilance could lead to increased heart rates and metabolic activity, subsequently adding other factors that possibly contribute to the negative energy balance of bears while on land.

Another bear–human interaction occurs in the form of the Polar Bear Alert Program (PBAP) at Churchill. The Manitoba provincial management agency initiated the program in 1969to protect local residents from bears, and vice versa (Kearney, 1989). The area around the town is patrolled, and bears that enter certain zones will be deterred, captured, handled, or destroyed. From its inception up to 2000, an average of 48bears per year (a total of 1547 bears) have been handled (Kearney, 1989; Calvert et al., 1991b, 1995b; Lunn et al., 1998; for a detailed PBAP description, see Kearney, 1989). Handling procedures are similar to those during research activities, and effects can be assumed to be similar.

Considering CWS-related research activities and the PBA Pactivities between 1977 and 1995, a total of 3558 bears (not including university-research handled bears) have been handled (last column in Table 1). This is about three times greater than the actual estimated WH population of 1100(Derocher and Stirling, 1992), indicating that all bears are, average, subject to repeated handling. Moreover, these activities occur when bears are either fasting or leaving their dens and are already energetically stressed. It is plausible that these repeated bear–human interactions have adversely stressed the bears over the past 30 years.

3. Food availability and competition

Between 1978 and 1990, the WH polar bear population was estimated to be around 1100 bears (Derocher and Stirling, 1992).Derocher and Stirling (1995) estimated the mean size of the population between 1978 and 1992 to be around 1000 bears. Up to 1997, the population did not change significantly, and was estimated to be around 1200 bears (Lunn et al., 1997a; Fig. 6 in Stirling et al., 1999). When published yearly population estimates from Derocher and Stirling (1995) and Lunn et al. (1997a) are examined, several tendencies are apparent. First, the Derocher and Stirling (1995) data for 1977–1992 show an increasing trend (F = 4.16, p = 0.06, r2 = 0.23), although that trend is not statistically significant. Second, the Lunn et al. (1997a) data from 1984 to 1995 indicate a stable population(F = 0.71, p = 0.42, r2 = 0.07).When both data sets are combined(i.e., the Derocher and Stirling (1995) data from1977 to 1992 and the Lunn et al. (1997a) data for 1993–1995), a significant increase in the population size is implied (F = 6.40, p = 0.02, r2 = 0.27).Most recently, however, it was noted that the population since1995 has been declining to ‘‘less than 950 in 2004’’ (IUCN/Polar Bear Specialist Group, 2005). We clarify that the published estimate by Lunn et al. (1997a), combining Churchill and Cape Tatnam study area (both in WH) datasets, gives a 1995WHpolarbear population of 1233 with a 95% confidence interval that ranges from 823 to 1643 bears, so the actual confidence in the ‘‘decline’’ of theWHpolarbearpopulationin2004, relative to the 1995 values, is difficult to confirm.

Given these long-term data on population estimates and responses, it is possible that density-dependent processes have been imprinted in the observed records of polar bearcat. It is important, however, to recognize the great difficulties in demonstrating density dependence in population studies (e.g., Ray and Hastings, 1996; Mayor and Schaefer, 2005), among which is the sensitivity of the phenomenon on spatial scale covered by the population sampling techniques (e.g., Tayloret al., 2001). We concur with Derocher and Stirling (1995) and Stirling et al. (2004) that the WH population was at least stable during the 1984–1995 period (and likely up to 1997; see Stirlinget al., 1999, their Fig. 6). Prior to that the WH population was hunted heavily, which led to hunting restrictions (Stirling et al., 1977; Derocher and Stirling, 1995). After the population recovered, and then increased, bear body mass, reproductive parameters, cub survival, and growth declined (Derocher and Stirling, 1992, 1998b). Derocher and Stirling (1992, 1995, 1998b) considered whether these responses reflect density dependent population control mechanisms. They discarded them either because no accurate population estimates for WH existed, or no change in population size was detected. Typically, density dependent responses, similar to those exhibited by WH polar bears, are detected in increasing populations (Eberhardt and Siniff, 1977; Fowler, 1990). By contrast, however, individuals of a population near carrying capacity (given that the WH

population remained relatively stable for so long)can also exhibit traits that were observed for this polar bear population, namely poorer physical condition, lower survivorship, and lower rates of reproduction (Kie et al., 1980, 2003; Stewart et al., 2005). It is possible that the WH population has been stable for so long because carrying capacity has been reached, and intraspecific competition increased with increasing polar bear density, resulting in the documented responses.

It is important to note that the southern half of Hudson Bay is shared between polar bear populations of WH and southern Hudson Bay (SH) (Derocher et al., 1998). Polar bears of SH have exhibited better body condition as compared to their WH counterparts (Stirling et al., 1999, 2004) but prolonged ice conditions in that area seem not to be the explanation because recent updated analysis by Gagnon and Gough (2005a) suggested tendencies toward earlier ice break-up (hence shorter overall duration of sea-ice cover) in James Bay and along the southern shore of Hudson Bay. Population estimates, which have been conducted almost entirely via aerial surveys, indicate an increasing trend for this SH population from1963 to1996 (i.e., see Table 2 and Fig. 4c of Stirling et al., 2004). Although both populations are recognized as independent (e.g., Derocherand Stirling, 1990a, b; Kolenosky et al., 1992; Taylor et al., 2001), possible overlap can occur on the sea-ice. If population density for SH has been increasing, whereas food supply has been insufficient due to increased competition, then some SH bears may have expanded their hunting forays, leading to competition for food with WH bears. Yet there has not been a drastic decline in the WH population detected. One reason may be that the bears have learned to hunt seals during the ice-free period along the shores in tidal flats. This phenomenon has been observed for several years at Churchill in the polar bear viewing area (Dyck, personal observations).

Data on the bear food supply is needed to draw more clear conclusions about the interplay between population densities and worsening physical attributes of polar bears. The main prey of polar bears are ringed (Phoca hispida) and bearded seals (Erignathus barbatus) (Stirling and Archibald, 1977; Smith, 1980), but seal population data are too limited at present to resolve this issue (Lunn et al., 1997b).

4. Air temperature and climate variability around Hudson Bay

Fig. 1a shows the surface air temperature records[2] of nearby Churchill, Manitoba (assumed here to be representative of WH) from 1932 to 2002 for the four climatological seasons. The large interannual variability of the seasonal temperatures suggests that establishing a meaningful long-term trend in any of these relatively short records would be difficult and that a trend determination, especially over short periods, will be

highly sensitive to the time interval considered (e.g., Pielkeet al., 2002; Cohen and Barlow, 2005). Fig. 1b attests that no statistically significant warming trend (dotted trend lines fitted over the full records in Fig. 1b) can be confirmed for either the late spring (defined here as the average of April, May and June, following discussion in Stirling et al., 1999) or fall seasons when the full record from 1932 through 2002 is considered. Thus, the hypothesis that a warming trend is the principal causative agent for the supposed earlier spring melt and later fall freeze of the sea-ice around WH cannot be confirmed. Further, that the temperature trend is not statistically different from zero indicates it is not obviously forced by anthropogenic greenhouse gases as commonly assumed and extrapolated to suggest implications for polar bear ecology in future scenarios of climate change. Such extrapolations remain premature at best.

An apparent tendency towards late spring warming can be derived by examining the period from 1981 to 1999, illustrated by the dashed trend curve in Fig. 1b. Clearly, the choice of endpoints is very influential on the results. The trend fails to persist when data through 2002 are included and we make no inferences about any concurrent ecological responses. Thus, although our independent results for temperature change and variability over the WH do not contradict Stirling et al. (1999) for the limited period from 1981 to 1999, the longer record reveals a fuller range of air temperature variability that argues against assuming a persistent warming trend.

Gough et al. (2004) recently identified snow depth as the primary governing parameter for the interannual variability of winter sea-ice thickness in Hudson Bay because of its direct insulating effect on ice surfaces. By contrast, the concurrent winter or previous summer air temperatures yield only weak statistical correlations with ice thickness. Detailed high resolution modeling efforts by Saucier et al. (2004) that considers tides, river runoff and daily meteorological forcing, found tidal mixing to be critically important for ice-ocean circulation within, and hence the regional climate of, the Hudson Bay basin.

We further examined records of winter and spring air temperatures at Frobisher Bay (now called Iqaluit, Nunavut) by the Hudson Strait and the respective winter and spring Arctic Oscillation (AO) circulation indices[3] (Fig. 2) to better characterize the regional pattern of air temperature variability. Fig. 2 shows two important points. First, note the rather strong cooling trend (at a rate of about 0.4 8C per decade since the 1950s) for the winter and spring temperatures of Frobisher Bay. Regional differences in the pattern of the temperature variability, especially on the multidecadal timescale, are large. This pattern of large temperature gradients between the southwestern and northeastern corners of the Hudson Bay oceanic basin has been well noted by Ball (1995), Catchpole (1995) and Skinner et al. (1998)—these authors also provided a comprehensive discussion on climate regimes around WH, including a broad, historical perspective on the range of natural variabilities. Among other things this indicates that a hypothesis of late spring warming negatively affecting the WH polar bear population ecology cannot be universally extended to other locations.

The second point of Fig. 2 is that the air temperature and climatic conditions around the Hudson Strait and Hudson Bay areas have a close association with the AO circulation index. The correlations shown in Fig. 2 are statistically significant, with AO variability explaining up to 20–50% of the interannual temperature variances at Frobisher Bay.

To examine the link between the AO index and Frobisher Bay air temperatures, both series were regressed on a matrix of monthly dummy variables to remove fixed seasonal effects. The residuals of these regressions (denoted AOrand FRr) were then tested in a vector auto regression to determine leading patterns of Granger causality (see Hamilton,1994, Chapter 11).While AOr shows a significant Granger causal pattern on FRr no such pattern exists in the other direction. This means the current value of the AO index significantly improves forecasts of monthly Frobisher Bay air temperatures, but the current air temperature does not improve forecasts of the AO. Finally, FRr was regressed on its first two lags, AOr and the first three lags of AOr to remove serial correlation in the mean. After a trend term that was insignificant was removed, the r2 from this regression was0.39 (with an adj-r2 of 0.38). A Wald test of the joint AOr terms yielded a chi-square (d.f. = 4) statistic of 235.6. A p-value on the hypothesis of no influence of the current and lagged AO anomalies on the current monthly temperature anomaly is less than 0.00001.

The AO circulation index appears to be physically relevant for two reasons. First, from an examination of the statistics of sea level pressure and sea-ice motion from the 1979 to1998 data collected by the International Arctic Buoy Programme, Rigor et al. (2002)confirmed that the AO circulation pattern can explain at least part of the thinning sea-ice trend observed over the Arctic Ocean. Polyakov and Johnson (2000) and Polyakov et al. (2003a) further emphasized the importance of the relative phasing of the decadal and multidecadal (i.e., 50–80 years) oscillatory modes of Arctic atmospheric circulation variability in explaining the recent Arctic sea-ice areal extent and thickness trends. Rigor et al. (2002) clarified that instead of assuming that the warming trend in surface air temperature caused the sea-ice to thin, it is the AO-induced circulation pattern that produces the tendencies for sea-ice to thin and sea-ice area to retreat (see further discussion on regional sea-ice trends and mechanisms in Zhang et al., 2000; Kimura and Wakatsuchi, 2001; Polyakov et al., 2003b; So¨ derkvist and Bjo¨ rk, 2004). In turn, it was the changes in sea-ice that caused the air temperature to warm because of an increasing heat flux from the interface with the ice-free ocean. Beyond atmospheric AO, Shimada et al. (2006) recently documented and highlighted the key role played by the inflows of warm Pacific summer water through the Bering Straits in causing the large sea-ice areal reduction in the Arctic that began in the late 1990s. Thus, such a complex physical picture connecting oceanic and atmospheric processes with sea-ice variability is dramatically different from Stirling et al. (1999)’s suggestion in which warm spring air temperature is considered to be the ultimate cause for the earlier spring sea-icebreak-up[4] and poorer conditions of polar bears.

The second reason to discuss the AO index is related to a recent finding that climatic change effects associated with the AO index are propagated through two trophic levels within a high-arctic ecosystem (Aanes et al., 2002). From the statistical analyses of the 1987–1998 growth series of Cassiope tetragona (Lapland Cassiope) and the 1978–1998 abundance series of an introduced Svalbard reindeer (Rangifer tarandus platyrhynchus) population near Broggerhalvoya, on the NW coast of Svalbard, Aanes et al. (2002) found that high positive values of the AO index are associated with decreased plant growth and reindeer population growth rate. Thus, the reindeer population at Svalbard, through the mediation of the climate modulated effects on plant growth, is plausibly connected to climate through a bottom-up sequence. But Aanes et al.(2002)noted that the bottom-up scenario may be density dependent in that at higher reindeer densities, a reverse top down sequence of trophic interaction is becoming more important in which grazing has a dominating influence on the forage species and plant communities. The AO index is thus promising as a useful climatic variable for further examination of the dynamic of trophic interactions under various settings of the arctic ecosystem.

It must also be asked whether natural climate oscillations as those described above – reducing sea-ice cover and changing the freeze-and-thaw cycles that affect the food sources of polar bears at higher latitudes – are really as detrimental to biodiversity as suggested. These changes may create more polynyas, which are productive oases in the ice (Stirling, 1997), or increase marine productivity overall (Fortieret al., 1996; Rysgaard et al., 1999; Hansen et al., 2003) primarily because of the modulation of the food web of the lower trophic levels by freshwater-limiting and light-limiting processes. Bears do not feed year-round, but do feed during late spring when seal pups are abundant. More fat deposits may be accumulated during this time, and a ‘‘true hibernation state’’ like black (U. americanus) and brown bears (U. arctos) could become an evolutionary strategy for the remainder of the year for polar bears. This scenario could be very likely because polar bears evolved from brown bears (Kurte´n, 1964). Alternatively, a supplementary feeding strategy could evolve where berries and vegetation are consumed in higher frequencies during the ice-free period, as has been observed for bears of Hudson Bay (Russell, 1975; Derocher et al., 1993).