Skip to main content
Dryad logo

Polar bears are inefficient predators of seabird eggs

Citation

Jagielski, Patrick (2021), Polar bears are inefficient predators of seabird eggs , Dryad, Dataset, https://doi.org/10.5061/dryad.prr4xgxjw

Abstract

Climate-mediated sea-ice loss is disrupting the foraging ecology of polar bears (Ursus maritimus) across much of their range. As a result, there have been increased reports of polar bears foraging on seabird eggs across parts of their range. Given that polar bears have evolved to hunt seals on ice, they may not be efficient predators of seabird eggs. We investigated bears’ foraging performance on common eider (Somateria mollissima) eggs on Mitivik Island, Nunavut, Canada to test whether bear decision-making heuristics are consistent with expectations of optimal foraging theory. Using aerial-drones, we recorded multiple foraging bouts over eleven days, and found that as clutches were depleted to completion, bears did not exhibit foraging behaviours matched to resource density. Bears visited fewer nests overall as the season progressed, but marginally increased their visitation to nests that were already empty. Bears did not display different movement modes related to nest density, but became less selective in their choice of clutches to consume. Lastly, bears that capitalized on visual cues of flushing eider hens significantly increased the number of clutches they consumed; however, they did not use this strategy consistently or universally. The foraging behaviours exhibited by polar bears in this study suggest they are inefficient predators of seabird eggs, particularly in the context of matching behaviours to resource density.  

Methods

Polar bear observations

We recorded foraging bears with drones (DJI Phantom 3 Pro and 4 Pro models, www.dji.com) from July 10-20, 2017 between 0530 hrs and 2030 hrs. We initiated filming when conditions were suitable for flying and when bears were actively foraging on eggs. Bears were filmed as soon as researchers noticed them on the island and bears were recorded until they either left the island or were resting for extended periods of time. The drone pilot and observer were stationed on the roof of a research cabin and launched/landed the drone within an electrified fence surrounding the research station. Drones were flown above the focal bear at altitudes that elicited minimal apparent behavioural responses (we suspect that any minimal behavioural response was also due to ambient noise of the colony, and bears continually being harassed by herring gulls (Larus argentatus); Figure 2a), although we cannot say with absolute certainty that bears and eiders were not affected physiologically (i.e., changes in heart rate [51, 52]). Videos were recorded at 30 frames per second at a resolution of 2700 x 1520 pixels. In total, we recorded 995 minutes of polar bear foraging footage. Video data were binned into distinct ‘foraging bouts’, which represented a near continuous observation of a bear foraging. Whenever there was a considerable time gap (mean = 167 minutes; median = 62 minutes; range = 426 minutes) in filming a focal animal, due to i) a change in bears’ activity (e.g., swimming or resting) which affected the continuity of a foraging bout, and/or ii) as a result of having to replace batteries due to flight time limitations of the drones used in this study (i.e., approximately 20 minutes), we considered it a new foraging bout. Our dataset consists of a maximum of 20 individual bears with 31 distinct foraging events (as some bears foraged more than once in a day) that range from 2.85 to 134 minutes (mean = 32 minutes; median = 26 minutes; range = 131 minutes). For further details on drone operations, including how we minimized bear disturbance, see the Drone Reporting Protocol [53] in supplementary materials.

Whenever possible, we distinguished individual bears by using: i) the date and time of video filming, ii) field logs containing information such as number of bears on the island on a particular day and time, and/or iii) the stains on bears’ fur and scars on bears’ faces/body. However, when a bear could not be distinguished from other individuals foraging in the colony during the same time period, we considered it a new individual. Bears were also not differentiated between days as we had no definitive way of distinguishing them with absolute certainty. Importantly, our intent was to capture polar bear foraging behaviours as the resource depleted, which is the main driving force affecting behaviours universally according to OFT, thus likely swamping out any individual variation.

Behavioural analyses

We used Solomon Coder version: beta 17.03.22 (https://solomoncoder.com), a manual behavioural coding tool, to analyse polar bear foraging behaviour. We loaded drone video data into this program and categorized bear behaviour (during video playback) based on predefined behaviours of interest. In reviewing each foraging event, we recorded: 1) “Empty nests visited”: number of empty nests a bear was considered to have visited when it was clearly evident from the video footage there were no eggs in the nest cup (see supplementary material: video 1a and Figure 2a). 2) “Movement sinuosity”: number of turns a bear made when searching for nests. We considered it a turn only when it was clearly visible that a bear abruptly (i.e., in 1-5 bear strides) veered a minimum of 45 degrees from its heading during locomotion or changed directions (i.e., turned left, right, or 180 degrees) after briefly stopping. We did not consider it a turn when a bear gradually (i.e., in > 5 strides) moved in another direction (see supplementary material: video 2 and Figure 2b). 3) “Nests ignored”: number of nests a bear was considered to have ignored when it walked up to a full clutch and did not consume any eggs (see supplementary material: video 1b and Figure 2c). When it was clearly evident that a bear inspected a nest (i.e., looked into the nest, sniffed the clutch) and then ignored it, we concluded that this was due to a bear choosing not to consume the eggs. This may possibly be due to bird feces on the eggs as a result of the eiders’ defence mechanism when flushing from their nests [54]. Therefore, for the purposes of this study, we assumed nests were ‘ignored’ due to a bear’s reduced preference for that particular clutch of eggs. 4) “Visual cues ignored and used”: number of visual cues a bear ignored (i.e., eider hen(s) flushing from nest) when it was clear that the focal bear had observed at least one duck flush (i.e., head was facing in the same direction as flushing hen(s)), and did not approach the newly abandoned clutch. Conversely, a bear was considered to have used a visual cue when it was clearly evident that it had observed at least one duck flush (i.e., head was facing in the same direction as flushing hen(s)), then switched its current heading to approach the exposed nest. Ducks flushing in the bears’ peripheries were not considered in this analysis (see supplementary material: videos 1c; 3a and Figure 2d). 5) “Clutches eaten”: number of clutches a bear was considered to have eaten when it was clearly observable that it was chewing, licking, and/or egg contents were dripping when its face was in the nest (see supplementary material: video 3b and Figure 2c). In addition, to further confirm an eating event, a full clutch had to be clearly visible upon approach, and/or a hen was seen flushing from the nest when the bear approached. Any approach to a nest not fulfilling the above criteria was considered an ‘empty nest visit’. Lastly, we analysed 6) “Total number of nests visited”: which was the sum of clutches eaten, empty nests visited, and nests ignored. This analysis is meant to contextualize the number of empty nests bears visited as the eider breeding season progressed. For example, should the number of empty nests encountered decline whilst bears increased total number of nests visited, the implications of this behaviour (increased foraging efficiency) would be different should the opposite trend be observed (decreased foraging efficiency).  In addition to these foraging behaviours, we also recorded the duration of time a bear spent walking and standing (i.e., searching), as well as the duration of time a bear spent foraging: ingesting eggs while either standing, sitting, laying down, or walking.

Statistical analyses 

We used generalized linear models (GLMs) and accounted for overdispersion found in the data by using negative binomial error distributions. Each of our models tests different components of foraging efficiency. Using the following response variables, we tested: discrimination using a) the total number of nests visited and contrasted it with b) the number of empty-nest visits; movement patterns with c) the number of turns bears made (i.e., movement sinuosity); selectivity using d) the number of nests ignored; and utility of visual cues by analysing e) number of clutches eaten. For variables a-d, we used foraging bout order as a continuous predictor variable (i.e., first recorded foraging bout = 1, last recorded foraging bout = 31) which serves as a proxy for resource density (since bears are continually consuming nests). For variable e, we used proportion of cues used (i.e., cues used divided by total cues observed) as our predictor variable, which encompasses the entire suite of events when a cue(s) was present and available for a bear to use. This serves to test whether using visual cues enhances bears’ ability to locate a greater number of nests. For all our models, we added search time (i.e., time spent walking and standing for each bear’s foraging event) as a fixed effect to account for differences in filmed-video lengths. See Table 1 for model overviews.

All statistical analyses were performed in R version 3.4.4 [55] using the tidyverse [56] and glmmTMB [57] packages. Figures were created using the ggplot2 [58], ggeffect [59], cowplot [60], and gridExtra [61] packages.