Den phenology and reproductive success of polar bears in a changing climate

Abstract

Synchrony between reproduction and food availability is important in mammals due to the high energetic costs of gestation and lactation. Female polar bears (Ursus maritimus) must accumulate sufficient energy reserves during spring through autumn to produce and nurse cubs during the winter months in snow dens. Adequate time in a den is important to optimize cub development for withstanding harsh Arctic spring conditions and to synchronize emergence with peak prey availability, which occurs in May and June. During 1985–2013, den phenology was investigated using temperature data collected on satellite collars deployed on adult female polar bears in the southern Beaufort Sea (SB) and Chukchi Sea (CS). We examined relationships between den phenology, reproductive success (cub production and post-emergence survival), and environmental factors (weather and sea-ice conditions). Females observed with cubs emerged later and remained in dens on average 15.0 ± 7.6 (SE) days longer than females seen without cubs. Females occupying land-based dens, where estimated snowfall was greater, had higher reproductive success. Recently, female polar bears have increased land-based denning in the SB. Females in CS emerged later from dens than SB females, consistent with better female body condition and higher cub survival in the CS. During years with a greater area of autumn sea ice, reproductive success was higher at land-based versus sea-ice dens, suggesting continued decline in sea ice could negatively affect recruitment. However, further research is needed to better understand mechanistic relationships. Because females emerging later from dens had higher reproductive success, den duration could be a useful metric in population monitoring.

habitat, polar bear, reproduction, snow, temperature, ursid, Ursus, maritimus

Climate change has been associated with changes in the phenology of numerous species (Parmesan and Yohe 2003). Warming temperatures are correlated with advances in spring events, particularly at higher latitudes, with examples that include earlier budding and flowering of plants, and changes in timing of hibernation, migration, and breeding (Parmesan 2006; Johnson et al. 2017). Temporal changes in important biological functions can disrupt tight trophic coupling (i.e., mutualistic or predator–prey interactions) when resources become available at the wrong time or place (Sydeman and Bograd 2009). Such trophic asynchrony can have population-level effects by reducing foraging efficiency, ultimately affecting reproductive success and survival (Durant et al. 2007; Sydeman and Bograd 2009).

Bears in Arctic and temperate regions are obligate seasonal breeders that typically mate in the spring and summer and produce cubs in dens during the winter months (Spady et al. 2007). During winter denning, females do not eat or drink, and survive solely on stored energy while undergoing delayed implantation, gestation, parturition, and lactation (Nelson et al. 1983; Spady et al. 2007). Cubs are born as small, altricial neonates and dens provide shelter during cub growth (Ramsay and Dunbrack 1986; Messier et al. 1994). Food availability for females prior to denning and after den emergence appears to affect cub production (Robbins et al. 2012) and survival of cubs after den emergence. Females that give birth to cubs need to replenish lost energy reserves to support lactation (Elowe and Dodge 1989; McDonald and Fuller 2005). Thus, females have to obtain sufficient food during the autumn to survive and produce cubs during the winter denning period (i.e., not enter the den too early and run out of fat reserves). In addition, females cannot emerge from the den before cubs have sufficiently developed and adequate food resources are available.

Pregnant female polar bears create dens in snow that provide a sheltered environment from harsh Arctic conditions (Ramsay and Stirling 1988). Because females rely on energy reserves to support cub production and growth, autumn body condition of pregnant females is an important factor affecting cub size, litter size, and subsequent survival (Derocher and Stirling 1996; Robbins et al. 2012). Pregnant females in poor condition (< 20% body fat) cannot produce cubs (Robbins et al. 2012), but previous studies suggested that the threshold for reproductive failure is low and that most females that enter dens produce cubs (Derocher et al. 1994). Growth rate of cubs is related to their mother’s condition (Robbins et al. 2012) and body mass of cubs is a predictor of cub survival (Derocher and Stirling 1996). Polar bear cubs increase their body mass by up to 4 times between spring and autumn and nursing provides the majority of calories for growth (Derocher and Stirling 1998). In western Hudson Bay, 25% of spring-captured females lost their entire litter by the autumn. The loss of the entire litter post-emergence accounted for the majority of females observed without cubs in the autumn. Effects of limited food resources during the autumn prior to and the spring following denning on maternal milk quantity and quality appear to be the primary cause of mortality in juvenile polar bears (Derocher et al. 1993; Derocher and Stirling 1996). Thus, a female’s body condition at entry into a den is a critical factor determining whether she produces cubs and their chance of survival during their first year.

The duration a female spends in the den may also affect cub survival and be dependent on food availability prior to den entrance. Brown (Ursus arctos) and black bears (U. americanus) delay entry to a den when food is abundant in the autumn, presumably to maximize energy reserves prior to denning (Van Daele et al. 1990; Schooley et al. 1994; Friebe et al. 2014; Pigeon et al. 2016; Johnson et al. 2017). Further, parturient female brown and black bears tend to emerge later from dens than males and non-reproductive females (Johnson and Pelton 1980; Garrison et al. 2007; Waller et al. 2012; Friebe et al. 2014; Johnson et al. 2017), suggesting that extended denning supports cub development. Polar bears emerge from dens between March and May, which coincides with a period of increasing prey availability; ringed seals (Phoca hispida) begin pupping and ringed and bearded seals (Erignathus barbatus) haul-out on the ice to molt during this period (Stirling and Archibald 1977; Smith 1980, 1987). Derocher and Stirling (1996) suggested that polar bear mothers and cubs depart for sea ice, instead of nursing at the den site, to obtain necessary food resources in the spring. Larger, more developed cubs are less susceptible to heat loss (Blix and Lentfer 1979) and may have improved abilities to travel on sea ice. Thus, cub survival may be maximized for females that obtain the greatest energy reserves prior to entering a den and therefore can remain in dens longer, allowing cubs to fully develop, and maximizing overlap with spring food availability.

The effects declining extent of sea ice (Durner et al. 2009; Laidre et al. 2015) and changes in snowfall patterns on land and sea ice (Webster et al. 2014) may have on the timing and duration of denning are not well understood. Polar bears in several regions, including the Chukchi Sea (CS) and southern Beaufort Sea (SB) and western Hudson Bay, spend more time on land during the months preceding denning than they have in the past (Rode et al. 2015b; Atwood et al. 2016), which can affect female body condition prior to denning (Stirling et al. 1999). On land, polar bears primarily rest (Whiteman et al. 2015; Ware et al. 2017) and feed minimally on terrestrial food resources (Rode et al. 2015a). Those bears that remain with the sea ice during the Arctic sea-ice minimum (prior to denning) exhibit declines in activity and appear to have reduced access to prey (Whiteman et al. 2015; Ware et al. 2017). Links between availability of sea ice in autumn, female body condition, and cub survival were documented in western Hudson Bay, but are less clear in other populations where monitoring of female body condition prior to den entry is less feasible. Bromaghin et al. (2015) identified reduced cub survival as a factor affecting population decline in the SB. Polar bears have also changed their denning behavior in some areas, including selection of den locations at higher latitudes (Derocher et al. 2011; Rode et al. 2015b) and elevations (Escajeda 2016), increased selection of land-based dens (Fischbach et al. 2007; Rode et al. 2015b; Olson et al. 2017), and changes in the time of entry to the den (Escajeda 2016). The implications of these changes on cub production and survival are unclear.

We examined relationships between the timing and duration of maternal denning and reproductive success (i.e., whether a female produced cubs and whether they survived for the first 100 days following den emergence) in the CS and SB subpopulations over 28 years. In these 2 subpopulations, females denned both on the sea ice and on land. Information on female body condition just prior to denning is limited, but bears in both subpopulations have altered habitat use during the months preceding denning (Rode et al. 2015b; Atwood et al. 2016). We hypothesized that a shorter denning period was associated with lower reproductive success in polar bears, either because a female did not produce cubs or she left the den with cubs that were less developed and thereby, less vigorous for surviving harsh Arctic spring conditions. Further, we hypothesized that a female’s denning period was related to food availability in autumn, which affected her body condition at the time of den entry and the amount of energy reserves available for the denning period. We sought to understand the factors that may influence the timing and duration of denning, including whether a bear denned on land or sea ice, and relationships with weather and sea-ice conditions. Because polar bears den in snow, we hypothesized that snowfall could be a factor affecting the timing and duration of denning. Lastly, because female body condition, cub survival, and recruitment differed between the 2 subpopulations (Rode et al. 2014) and because summer habitat use and denning sites have changed in these subpopulations (Fischbach et al. 2007; Rode et al. 2015b; Olson et al. 2017), we hypothesized that the timing and duration of denning was associated with the observed differences in reproductive success.

Materials and Methods

Satellite radiocollar location data have been used to identify denning behavior in brown and black bears (Ciarniello et al. 2005; Johnson et al. 2017), but because some female polar bears den on the sea ice, ice movement precludes identifying denning behavior using satellite radiocollar location data. Therefore, we estimated the timing and duration of denning using air temperature-sensor data collected from collared female polar bears as described below. Temperature data from collars of denning bears exhibit distinguishable, prolonged increases compared to the colder, more fluctuating ambient temperatures recorded on collars of non-denning bears (Olson et al. 2017). Estimates of the dates of den entry and emergence using temperature data (as described in greater detail below) likely represent the time frame that a polar bear is consistently in a den rather than the time frame a bear is at a den location, which has been described for other bear species that den on stable habitats (Ciarniello et al. 2005; Johnson et al. 2017). Because temperature has not been used previously to identify the timing and duration of denning, we compared temperature-based estimates of the denning period with location-based estimates for polar bears that denned on land, where sea-ice drift would not affect location data. The results of this comparison were used to assess the accuracy of our temperature-based approach recognizing that a bear may occur at a den site and not be in the den.

Bears were captured on the sea ice during spring (mid-March to mid-May) and occasionally during autumn (August to November) in the Alaska portion of the SB during 1985–2013, with the exception of 1993, 1995, 1996, and 2010. Captures occurred in the CS and northern Bering Sea during 1986–1995 and during 2008–2013 on or near Wrangel Island, Herald Island, the Alaskan mainland coast, St. Lawrence Island, Alaska, and the northeast Chukotkan coast. Body condition just prior to den entry was not known for most females because most were captured and collared during the previous spring. Bears were located via helicopter, immobilized, and adult females were fitted with satellite radiocollars (Telonics, Inc., Mesa, Arizona). Immobilization procedures and collar deployments are further described in Rode et al. 2015b and Olson et al. (2017). The methods for capture and handling bears used in this study conformed to the guidelines of the American Society of Mammalogists for the use of wild mammals in research (Sikes et al. 2016), were conducted under USFWS research permits MA 690038 and 04608, and followed protocols approved by Animal Care and Use Committees of the USGS and the USFWS.

Location data from radiocollars were used to assign bears to a subpopulation. Satellite radiocollars (Telonics, Inc.) provided either Argos (www.argos-system.org) or Global Positioning System (GPS) locations every hour to every 5 days. Location data were filtered to remove implausible locations using the Douglas Argos-Filter algorithm, which retained all standard quality class locations (classes 3, 2, and 1; < 1,500 m error), rejected all class Z locations, and retained class A and B locations if they were corroborated by a consecutive location within 10 km, or if movement rates were < 10 km/h and turning angles were not extremely acute (Rode et al. 2015b). Bears were assigned to a subpopulation based on whether the majority of their locations occurred within the SB or CS subpopulation boundaries as defined by the International Union for the Conservation of Nature’s Polar Bear Specialist Group (Obbard et al. 2010; Ware et al. 2017; Fig. 1).

Fig. 1.

Locations of maternal dens of polar bears (Ursus maritimus) identified using temperature-sensor data collected from satellite radiocollars deployed on adult female polar bears in the Chukchi Sea and Beaufort Sea subpopulations. Red lines indicate IUCN Polar Bear Specialist Group identified subpopulation boundaries.

Temperature data were recorded every 20 min by a thermistor integrated into a GPS radiocollar or transmitted once per 4 to 7 h daily duty cycle for Argos collars. The mean number of temperatures taken per day of data acquisition was 5.5 ± 0.03 (SE). However, the frequency of temperature data often was reduced once a bear entered a den due to attenuation of the signal required for satellite reception. Dens were built in the snow regardless of whether a female denned on land or sea ice. Thus, signal attenuation was expected to be similar on these 2 substrates.

Estimating den timing.—

We used statistical methods to identify maternal denning behavior (as described in Olson et al. 2017), entry into dens, emergence from dens, and denning duration using air temperature-sensor data from polar bear collars. This process quantified the expected mean and variation in temperature for non-denning bears, and identified bears as denning when air temperatures rose approximately 9°C above this expected range (referred to as a control limit—Olson et al. 2017) for a prolonged period (Fig. 2). When compared to direct observations of denning behavior, this temperature-based approach accurately classified denning and non-denning behavior in 94.5% of 73 bears (Olson et al. 2017). Per the methods of Olson et al. (2017), we excluded potential denning events that lasted less than 34 days, which could have eliminated some bears that entered dens but did not produce cubs, or abandoned the den early. We assumed that all denning events were likely maternal denning attempts because shelter denning (i.e., denning for reasons other than to give birth to cubs) is rare in the CS and SB (Olson et al. 2017). Entrance into a den was defined as the median date between the last temperature observation within control limits and the first observation above control limits. Similarly, emergence from a den was estimated as the median date between the last observation above control limits and the first observation to return within control limits. The total den duration was defined as the number of days between entrance into and emergence from a den (Fig. 2). This value most likely represented the time that a bear remained consistently within a den rather than the total amount of time a bear could have spent at the den location in preparation for or subsequent to denning.

Fig. 2.

Examples of control charts with temperature data measured by thermistors onboard satellite collars fitted to adult female polar bears (Ursus maritimus) used to identify denning and to estimate entrance into and emergence from dens. Entrance and emergence dates were estimated as the median date between observations within and above control limits at the start and end of a denning event (shown as horizontal dashed lines).

Dens were assigned as occurring on land or sea ice using GPS and Argos locations as described in Olson et al. (2017). Location data were filtered as described above, and if at least 1 observed location at the start of, during, or at the end of the denning period identified via temperature data occurred on land and locations preceding or subsequent to denning demonstrated a trajectory to or from that location, we assumed the den occurred on land (Olson et al. 2017). Dens that did not occur on land were designated as sea-ice dens.

Assessing the accuracy of temperature-based denning estimates.—

Temperature-based estimates of entrance into a den, emergence from a den, and denning duration were compared to location-based estimates using satellite collar location data of bears that denned on land and provided location data at least every 3 days. Bears that denned on sea ice could not be included in this comparison because ice drift precludes detection of denning behavior from location data. Location data were filtered as described above. Locations within a 1,500-m radius of the den location were queried, and we estimated the den entry and emergence dates as the first and last locations within that radius, respectively.

Assessing female reproductive success.—

Females that entered dens were considered reproductively successful if they were subsequently observed with dependent young within 100 days post-emergence (hereafter, “reproductive success”). Individuals that denned, but were observed without dependent cubs within 100 days following den emergence were considered unsuccessful. Differences in reproductive success were a consequence of both whether a female produced cubs and cub survival. Reproductive success was visually confirmed via VHF radiotracking or random encounter during ongoing mark-recapture efforts. Because mortality rates for cubs within their first year can be as high as 40–75% (Ramsay and Stirling 1988; Elowe and Dodge 1989; Wiig 1998), the time between denning and subsequent observation could have influenced observations of reproductive outcomes. Thus, we measured the number of days between den emergence and subsequent observation of females to determine potential effects on the observation of cubs and included these values in analyses of the factors influencing reproductive success.

Differences between subpopulations and trends over time.—

We were interested in whether the timing and duration of denning differed between subpopulations or over time. Because the proportion of females denning on land and sea ice differed between the 2 subpopulations, we included den substrate (0 = sea ice, 1 = land) as a fixed-effect covariate in models as described below (see “Statistical analysis”). Further, we were interested in whether the timing and duration of denning differed between substrates independent of subpopulations and trends over time, so we conducted direct comparisons between land and sea-ice dens. Finally, in a separate analysis described below, we examined whether the timing and duration of denning was related to spatial variation in weather and sea-ice conditions.

Environmental factors influencing denning.—

Polar bears require sea ice to access ice-associated seals, which are their primary prey (Thiemann et al. 2008). However, data on annual variation in seal abundance are lacking for these regions. Because the most significant loss of sea ice in the Arctic has occurred during the summer months (Stern and Laidre 2016), the extent of sea ice could affect predation success and deposition of energy reserves prior to denning. Therefore, we created an index of autumn food availability during the year each bear denned by estimating the amount of sea ice over the shallow continental shelf, the preferred habitat of polar bears (Durner et al. 2009). The percentage of the continental shelf covered by sea ice of ≥ 15% concentration during October and November (prior to den entry) was summarized using daily sea-ice concentration data from the National Snow and Ice Data Center (Cavalieri et al. 1996; accessed 16 October 2014). These values were annual measures applied to each bear during the years they denned and, therefore, were not specific to a bear’s den location. A similar metric was related to denning distribution (Rode et al. 2015b) and body condition (Rode et al. 2014).

Snowfall and air temperatures near the den have been associated with variability in the timing of denning in brown bears, despite denning occurring primarily in earthen dens (Zedrosser et al. 2006; Evans et al. 2016; Pigeon et al. 2016). Snowfall could be particularly important for polar bears because they den in the snow either on land or on the sea ice. Although the depth of snow required for polar bear denning is unknown and measurement of snow depth across the landscape is not possible, we examined whether an estimate of snowfall might relate to the duration and timing of denning because the amount of snowfall could affect the availability of adequate snow for denning. Therefore, we estimated snowfall and temperature at den locations using air temperature and precipitation data from the North American Regional Reanalysis (Mesinger et al. 2006), extracted using the Movebank Environmental Data Portal (Dodge et al. 2013). Weather variables were calculated as the inverse-distance weighted average among values from 4 NARR grid points (32 km resolution) surrounding each bear’s den location. We extracted air temperature and precipitation data at a level of 2 m a.s.l. and at 3-h intervals between October and March for each bear’s den location during the years in which they denned. For bear’s that denned on the sea ice, a location identified early in the denning period was used and the cumulative amount of precipitation was calculated when air temperature was below freezing (0°C). Between 1 October and 15 November of each year was designated as “autumn snowfall” and between 1 October and 1 March of each year was referred to as “spring snowfall.” The latter time frame was used to estimate accumulated snowfall prior to the period of emergence from dens. We assumed that precipitation below freezing would be in the form of snow rather than rain. The mean date of entry into dens was 15 November, and the mean date of emergence from dens was 1 March, thus we examined potential snowfall levels prior to these dates.

Statistical analysis.—

Paired t-tests were used to compare the dates of entrance and emergence from dens, and the total duration of denning estimated using collar temperature data with those estimated using location data of females that denned on land. This comparison was used to assess the accuracy of temperature-based estimates of the timing and duration of denning.

We compared temperature-based estimates of the duration and timing of denning between reproductively successful and unsuccessful females using logistic regression (0 = observed with cubs and 1 = not observed with cubs) with the number of days post-emergence when the observation occurred and den substrate as a covariate. In a separate logistic regression, we examined potential influences of weather on reproductive success because these variables were spatially and temporally explicit. Variables were considered to be collinear if condition indices were > 15, variance proportions > 0.3, tolerances < 0.1, and variance inflation factor > 10. If collinearity occurred, variables were included in separate models. We did not employ model selection based on reasons similar to those outlined in Hobbs et al. (2012) and the few variables in the model, but rather examined whether these hypothetical factors that could affect reproductive success were significant at P ≤ 0.05 and improved the model fit based on changes in the Akaike information criterion (AIC) value.

We used a general linear model (GLM) to compare entrance into dens, emergence from dens, and denning duration between the 2 subpopulations and to examine patterns over time. Models included year, population, den substrate (i.e., land or sea ice), and a year-by-population interaction. Den substrate was included to account for variability in the proportion of bears denning on land versus sea ice, which has changed over time (Rode et al. 2015b; Atwood et al. 2016). The interactive effect between population and year was included to allow for potential differences in trends over time between subpopulations. This effect was removed from the model if it was not significant and did not further reduce the AIC value of the model. We conducted an analysis of variance (ANOVA) to compare entrance into dens, emergence from dens, and denning duration between dens on sea ice and land. Finally, a GLM was used to examine potential effects of weather and sea-ice conditions on spatial and temporal variation in den entrance and emergence dates independent of potential variation over time, across den substrates, or between subpopulations (i.e., these variables were excluded from the model). Mean October air temperature, autumn snowfall, and October–November ice conditions were included in the model for entrance into dens and mean March air temperature, spring snowfall, and October–November ice conditions were included in the model for emergence from dens. Collinearity was examined and model selection was conducted as described above. Sea-ice conditions were included in models for both entrance and emergence because they were used as a proxy for autumn food availability, which could affect energy reserves of females and, thereby, both the timing of entrance and emergence. Because polar bears in this region have increased use of land during the summer for denning in recent years in relation to declines in summer and autumn availability of sea ice (Rode et al. 2015b; Atwood et al. 2016), these 2 factors were potentially correlated. Further, because den substrate is a binary variable, addressing this potential collinearity is complex. Therefore, the potential relationships between autumn sea-ice conditions and denning entrance, emergence, and duration were run separately for sea-ice and land-based dens. Further, to understand the role these variables might play in any observed variation in the timing and duration of denning across substrates, subpopulations, or over time, we compared weather and sea-ice conditions across these variables.

The Olson et al. (2017) algorithm for identifying denning behavior used SAS/STAT software. All other statistical analyses were performed using SPSS (IBM SPSS Statistics Version 24.0.0.0). All analyses were conducted at the alpha equals 0.05 level of significance.

Results

Assessing the accuracy of temperature-based denning estimates.—

Temperature-based estimates of the time of entrance into dens for land-based dens did not differ from estimates based on location data (paired t-test: t₃₀ = −0.36, P = 0.72). However, temperature-based estimates of the time of emergence from dens were 2.8 days earlier than estimates using location data (t₃₃ = 2.1, P < 0.0001), which could reflect the difference between emergence from dens and abandonment or be a consequence of the less-than-daily location data for some individuals. Location data were collected daily for 17 bears, every 2 days for 5 bears, and every 3 days for the remaining 9–12 bears in this analysis. All subsequent analyses described below used temperature-based estimates of the duration and timing of denning due to its accuracy for both land- and sea-ice-based dens.

Assessing reproductive success.—

Data on reproductive success were available for 72 bears, 8 of which were assigned to the CS subpopulation and 64 of which were assigned to the SB subpopulation. There was no difference in the mean date of entrance into dens between bears that successfully produced cubs (17 November; n = 50) and those that were observed without cubs (15 November; n = 21; F_1,69 = 0.05, P = 0.83). However, bears later observed with cubs (n = 50) emerged 14.9 ± 5.0 (⁠

$\overline{X}$

± SE) days later than females without cubs (n = 21; F_1,63 = 9.0, P = 0.004), and had an overall denning duration 15.0 ± 7.6 days longer than bears observed without cubs (F_1,63 = 3.9, P = 0.05). The earliest estimated date of emergence from a den for a female later observed with cubs was 8 January, with a duration of 61 days. The shortest estimated denning duration for a reproductively successful female was 42 days. The timing of observations post-emergence did not differ between females observed with cubs and those observed without cubs (logistic regression: β = 0.005 ± 0.009 days; P = 0.56). However, we retained this metric as a covariate in models of reproductive success.

Reproductive success was higher for females denning on land than on sea ice (logistic regression: β = 2.5 ± 0.8, P = 0.001; Fig. 3); however, when accounting for this difference there was no effect of autumn ice conditions (β = 0.02 ± 0.02, P = 0.30). No weather variables were related to reproductive success (all P-values > 0.05). Mean denning duration for females that produced cubs that survived until observation post-emergence was 113.8 ± 3.8 days (range: 42–157; n = 48), with post-emergence observations occurring within a mean of 37.4 days post-emergence. Mean denning duration of females that were later observed without cubs was 98.9 ± 7.4 days (range: 47–161; n = 21). Of females with den durations < 100 days, 43.8% were later observed with cubs compared to 78.2% of females with durations > 100 days.

Fig. 3.

Den locations for female polar bears (Ursus maritimus) that were observed with (dens that produced cubs) or without cubs (dens that did not produce cubs) following den emergence. Females were observed on average 37 days after emergence from dens.

Differences between subpopulations and trends over time.—

There were no differences in entrance dates for females in the 2 subpopulations (GLM with den substrate as a covariate: χ² = 0.13, P = 0.25; n = 215) and no trends among years (χ² = 0.02, P = 0.90; n = 215). Emergence from dens, however, was 9.4 ± 4.5 (SE; n = 50) days later in the CS compared to the SB (χ² = 4.3, P = 0.04; n = 129), but did not demonstrate a trend among years (χ² = 0.78, P = 0.38; n = 179). Mean date of entrance into dens was 15 November ± 1.9 days (n = 215) and mean date of emergence from dens was 1 March ± 2.1 days (n = 179). Denning duration also did not differ between subpopulations (χ² = 0.14, P = 0.71; n = 179) or among years (χ² = 0.94, P = 0.33; n = 179) and averaged 104.5 ± 2.8 days (n = 179). There was no interaction between subpopulation and year, or den substrate (land or sea ice) and subpopulation, in any of the models (all P > 0.2), so the interaction was not included.

Environmental factors influencing denning.—

Date of entrance into dens did not differ between land (n = 107) and sea-ice dens (F_1,191 = 1.16, P = 0.28; n = 108), but date of emergence was 8.7 days later (F_1,160 = 4.70, P = 0.03) and denning duration was 13.6 days longer for land-based dens compared to dens on sea ice (F_1,160 = 6.2, P = 0.014; n = 156). Weather and ice metrics were not collinear. Date of entrance into dens was positively related to mean temperature in October (β = 0.63 ± 0.23, χ² = 7.7, P = 0.006, n = 215), but not autumn snowfall (χ² = 0.51, P = 0.48) or availability of ice in October–November (χ² = 0.25, P = 0.62, n = 215). The model with the lowest AIC for den emergence included temperature in March (β = 0.42 ± 0.26; χ² = 2.6, P = 0.11, n = 156) and availability of ice in October–November (β = −0.16 ± 0.10, χ² = 2.5, P = 0.12, n = 156), but neither variable was significant in the model. Because changes in den substrate have occurred simultaneously with declines in availability of sea ice and therefore were potentially related, we examined relationships between autumn sea-ice conditions and time of denning for land- and sea-ice-based dens separately. There was no relationship between autumn sea-ice conditions and date of entrance into dens for land-based (β = −0.20 ± 0.13, χ2 = 2.1, P = 0.14, n = 89) or sea-ice-based dens (β = −0.11 ± 0.14, χ² = 56, P = 0.45; n = 73). Emergence date was later for land-based dens when there was greater availability of sea-ice habitat in October–November (β = −0.29 ± 0.12, χ² = 5.9, P = 0.02, n = 89), but there was no relationship for sea-ice-based dens (β = −0.20 ± 0.15, χ² = 1.9, P = 0.17, n = 73).

Environmental conditions differed between land- and ice-based dens with mean air temperatures in October 5.7°C warmer (F_1,183 = 24.9, P < 0.0001), autumn snowfall 12.3 cm greater (F_1,183 = 27.8, P < 0.0001), and spring snowfall 26.1 cm greater (F_1,183 = 29.4, P < 0.0001) at land-based versus sea-ice-based dens. There was no difference in temperatures in March between land- and sea-ice-based dens (F_1,183 = 1.9, P = 0.17). Conditions also varied between den locations in the CS and SB. Autumn snowfall was 10.1 cm greater (F_1,199 = 13.8, P < 0.0001), mean temperature in October was 5.2°C warmer (F_1,199 = 18.1, P < 0.0001), spring snowfall was 30.2 cm greater (F_1,199 = 44.2, P < 0.0001), and mean temperature in March was 7.1°C cooler (F_1,199 = 41.9, P < 0.0001) at den locations in CS compared to those in SB.

Discussion

Later emergence from the den by female polar bears in this study was most strongly associated with their likelihood of successfully producing and raising cubs within the first 100 days post-emergence. All of the females that denned through the end of March (12 of 65) were later observed with cubs, whereas approximately one-half of females that emerged prior to the end of February either never produced cubs or produced cubs that did not survive. Females that denned on land, where estimated snowfall during the weeks prior to and during denning was greatest, had higher reproductive success than females that denned on the sea ice. Because polar bears typically den in the snow rather than the earthen dens common to other bear species, snowfall at den locations could ensure the integrity of the den throughout the duration of the denning period. Further, many of the bears that den onshore in the SB spend the summer and early autumn onshore (Olson et al. 2017) where they have access to subsistence-harvested bowhead whales (Atwood et al. 2016), a unique, predictable food resource. In contrast, females that summer on the sea ice in the SB appear to have reduced access to prey (Whiteman et al. 2015) and some of these remain on the ice to den (Olson et al. 2017). Additionally, polar bears that summer on land appear to have earlier access to prime foraging habitats once the sea ice returns in the early autumn (Schliebe et al. 2008) compared to those that summer and den further north over deep water in the Arctic Basin. The relationship between sea-ice conditions in autumn and date of emergence from dens observed in this study are suggestive that availability of sea ice just prior to denning could contribute to reproductive success of polar bears denning on land, which represents the majority of denning females in these 2 subpopulations. In the CS, > 84% of females den on land and the proportion of land dens has not changed since the 1980s (Rode et al. 2015b). In the SB, land-based denning increased from 34.4% in 1985–1995 to 55.2% in 2007–2013 (Olson et al. 2017), a shift that could be associated with the apparently lower reproductive success of dens on the sea ice.

Food availability also affects the timing of den entry in black and brown bears with greater abundance of food linked to later entry into dens (Van Daele et al. 1990; Schooley et al. 1994; Friebe et al. 2014; Pigeon et al. 2016; Johnson et al. 2017). Although we attempted to use the extent of sea ice over the continental shelf as a proxy for food availability in autumn, there was no relationship between date of entry into dens and food availability, even when accounting for potential differences between land- and sea-ice-based dens. Changing sea-ice conditions over time, differences in potential access to food resources for the 2 subpopulations, different denning habitats (land versus sea ice), and the lack of data on seal abundance complicated the ability to detect whether food availability affected timing of den entry. Alternatively, because polar bears den exclusively to reproduce and not in response to reduced availability of food, this factor could be less important as a cue for denning in polar bears compared to bear species in temperate areas with significant seasonal fluctuations in food availability. Ambient air temperature also appeared to affect the timing of entry into dens, similar to other bear species in more temperate regions (Evans et al. 2016; Johnson et al. 2017). Our estimates of weather variables had coarse spatial resolutions that could have limited the ability to detect relationships with denning entrance, emergence, and duration (i.e., resulted in Type 1 errors). Further, the metric we used for sea-ice conditions was not spatially specific to the den location.

Our temperature-based estimates of entry into and emergence from dens followed expected patterns relative to estimates made using locations of land-based bears and were similar to previous observations made in this region (Amstrup and Gardner 1994). Emergence as identified by temperature-sensor data likely represents the opening of the den cavity and increasing exposure to ambient temperatures, whereas methods based on location data mark the departure from the den locale. Our result that emergence from the den occurred prior to den abandonment is consistent with reports that family groups remain at the den for a time before departing to foraging areas (Uspenski and Kistchinski 1972; Smith et al. 2007).

The length of the denning period varies substantially across populations of brown and black bears (80–197 days—Johnson and Pelton 1980; Schooley et al. 1994; Friebe et al. 2014), which could be the result of latitudinal variation in seasonal food availability (Spady et al. 2007). In contrast, denning in polar bears occurs only in females producing cubs and should be a function solely of the conditions required to maximize reproductive success. Bears and other species with delayed implantation, such as the European badger (Meles meles), vary the implantation date, with the fattest females typically implanting earliest allowing the longest growth period for young prior to emerging from the den (Woodroffe 1995; Robbins et al. 2012). Gestation in ursids lasts approximately 60 days (Tsubota et al. 1987; Spady et al. 2007) and females of the genus Ursus likely implant after they have entered a den (Kordek and Lindzey 1980; Tsubota and Kanagawa 1993). However, 2 of the 71 females in this study produced cubs after an estimated denning duration of 42 and 59 days, suggesting some flexibility in the den duration required to produce cubs. Although den duration was related to reproductive success, this relationship was weaker than that observed with emergence from dens. On average, denning duration of females that successfully reared cubs was 114 days, which was lower than the mean denning duration for other populations of polar bears (Messier et al. 1994: 186 days for females in the Canadian Arctic archipelago; Wiig 1998: 153 days for females at Svalbard; Escajeda 2016: 167–194 days in Baffin Bay and Kane Basin). The 2 Alaska subpopulations of polar bears we examined could be limited in denning duration by their inability to deposit sufficient fat reserves to support a longer denning period.

The SB subpopulation currently has some of the lowest cub survival rates in the Arctic (Regehr et al. 2017). Although SB polar bears only emerged 9 days earlier than CS bears and had similar denning durations, this difference could be biologically meaningful. SB females are in poorer condition in the springtime than CS females (Rode et al. 2014) and have lower availability of food during that time (Rode et al. 2017). Less is known about differences in body condition and feeding behavior in the summer and autumn, but lower food availability earlier in the year in the SB ecosystem could have cumulative effects on a female polar bear’s condition at den entry. Very low cub survival between 2003 and 2007 in the SB was identified as a key factor leading to declines in abundance for that population (Bromaghin et al. 2015). We did not detect a change in den entrance, emergence, or duration over time for either subpopulation in this study. However, 80% of our annual sample sizes were < 10 individuals and considerable variation in denning duration (range 33–219 days) could have limited detection of patterns to those that are large and broad-scale (e.g., sea-ice versus land dens). Further research is needed to understand annual trends in cub production and survival relative to the timing of denning.

Although polar bears in the SB have increased their use of land for denning since the 1980s (Fischbach et al. 2007; Olson et al. 2017), increased swimming has been documented in the summer and early autumn (Pagano et al. 2012). Increased distances between land and sea ice in the summer could impact accessibility to land habitats for denning as loss of sea ice continues (Bergen et al. 2007). Reduced access to denning habitat due to changes in sea ice was a factor contributing to lower density of maternal dens on Hopen Island in the southern portion of the Svalbard Archipelago (Derocher et al. 2011), and land-based habitats for denning, where reproductive success was highest, occur at the southern portion of the range of both subpopulations in this study. Polar bears could be challenged in coping with global warming because they have distinctly seasonal reproductive patterns and reproductive success that is heavily dependent on accumulated stores of body fat (Spady et al. 2007). Thereby, they could be susceptible to asynchrony between prey availability and the timing of denning, as well as direct effects of habitat loss on access to prey. Although reproductive success varied across a wide range of dates of emergence from dens, polar bears emerging later had much higher likelihoods of reproductive success, suggesting that this could be a useful metric to consider in population monitoring.

Acknowledgments

Studies were conducted under U.S. Fish and Wildlife Service research permit MA 690038 and followed protocols approved by Animal Care and Use Committees of the USGS (assurance no. 2010-3). Principal funding for this study was provided by the U.S. Geological Survey Ecosystems Mission Area and Changing Arctic Ecosystems Initiative and the U.S. Fish and Wildlife Service. Additional support was provided by BP Exploration Alaska, Inc., ARCO Alaska Inc., Conoco-Phillips, Inc., the ExxonMobil Production Company, Polar Bears International, Detroit Zoological Association, National Fish and Wildlife Foundation, a Coastal Impact Assessment Program grant through the State of Alaska, and Teck Alaska, Inc. This paper was reviewed and approved by USGS under their Fundamental Science Practices policy (http://www.usgs.gov/fsp). Use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Author notes

Present address of JO: U.S. Forest Service, Tonto National Forest, 2324 E. McDowell Road, Phoenix, AZ 85006, USA

Present address of EVR: University of Washington, Applied Physics Laboratory, Polar Science Center, Seattle, WA 98105-6698, USA