Bowhead whales overwinter in the Amundsen Gulf and Eastern Beaufort Sea

The bowhead whale is the only baleen whale endemic to the Arctic and is well adapted to this environment. Bowheads live near the polar ice edge for much of the year and although sea ice dynamics are not the only driver of their annual migratory movements, it likely plays a key role. Given the intrinsic variability of open water and ice, one might expect bowhead migratory plasticity to be high and linked to this proximate environmental factor. Here, through a network of underwater passive acoustic recorders, we document the first known occurrence of bowheads overwintering in what is normally their summer foraging grounds in the Amundsen Gulf and eastern Beaufort Sea. The underlying question is whether this is the leading edge of a phenological shift in a species' migratory behaviour in an environment undergoing dramatic shifts due to climate change.


Introduction
Climate change is predicted to alter animal migratory movements, primarily timing and range extensions [1,2], although the degree of phenotypic plasticity appears to be taxonomically specific [3][4][5]. Some seasonally migrating species are known to vary the extent and timing of their movements in relation to current environmental conditions [6,7] which may be related to differences in environmental variability as well as what is driving the migratory behaviour in the first place [5,8]. Of particular importance to the conservation of many species in the current context of climate change is the degree to which phenotypic plasticity related to migratory behaviour facilitates adaptation to a warming world [9].
Cetaceans, particularly baleen whales, are well known longdistance migrants, many species having some of the longest royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 8: 202268 several passive acoustic underwater recordings which confirm and provide more detail on these sightings, providing evidence of bowhead whales overwintering in the Beaufort Sea and Amundsen Gulf, a migratory anomaly that could be directly linked to climate change.

Acoustic recordings
We collected passive acoustic data at four locations in the south Amundsen Gulf during the 2018-2019 winter (figure 1) and compared these to several of our previous recordings in the region (see Comparisons with Other Datasets below). We deployed a Wildlife Acoustics SM3M acoustic recorder (Maynard, Massachusetts, USA) fitted with a low-noise hydrophone (High Tech, Inc., Gulfport, Mississippi, USA) between July 2018 and July 2019 near Ulukhaktok (70°42.857 0 N, 117°48.020 0 W). The acoustic recorder was set to a 48 kHz sample rate with 16-bit depth, a duty cycle of 5 min of recording every hour, and +12 dB of gain. The positively buoyant acoustic recorder was suspended roughly 3 m above the sea floor in 30 m of water and stopped recording on 24 April 2019 after running out of power. We also deployed three ST500 acoustic recorders (Ocean Instruments NZ

Bowhead whale acoustic detections
We processed all acoustic data using an automated detector and classifier for Arctic marine mammals (software: Spectro Detector; JASCO Applied Sciences Ltd, Victoria, BC, Canada), which we have used to successfully detect and classify bowhead whale vocalizations, and also beluga whales and bearded seals, in the western Canadian Arctic on multiple datasets [31][32][33][34]. The detector is fully described in Mouy et al. [35]. Briefly, the software detects acoustic signals between 10 Hz and 8 kHz, and extracts salient features for each signal, such as bandwidth and duration, calculated as in Fristrup & Watkins [36] and Mellinger & Bradbury [37]. A random forest classifier [38] is then used to assign each acoustic signal to a species. The classifier was trained on manually annotated passive acoustic data from the Chukchi Sea [35,39]. 'Manual' analysis refers to the process of aural and visual signal verification and categorization by a trained analyst, as compared to an automated process using classification algorithms. We performed manual bioacoustic analyses using Raven Pro Software (v. 1.5 and 1.6; Bioacoustics Research Program 2017) by manually inspecting spectrograms (window size: 7000 samples, DFT size: 8192 samples, overlap: 50%) with the time axis set to 15 s and the frequency axis from 0 to 3000 Hz, although the analyst could zoom in or out in either time or frequency as necessary. We manually verified all automated detections of bowhead whale vocalizations between 1 October 2018 and 15 April 2019 (24 April 2019 for the Ulukhaktok site), and manually analysed approximately 10% of files without bowhead detections (systematically selected as the 10th file in a row without a detection). Bowhead whale moans are identified based on their low-frequency characteristics (typically 50-300 Hz frequency range) and length (typically 0.5-2 s long, but can be up to 5 s long) [40,41]. No other species present in this region make vocalizations similar to bowhead whales [40,41], although occasionally sea ice can make low frequency, short tonal sounds that might confuse the automated detector and can sound similar to bowhead calls (W. Halliday, personal observation, 2020). For all recordings in January through April, we also assessed the file directly before and after any file with automated detections in order to increase our chances of finding all bowhead whale vocalizations recorded over the winter when bowheads are not expected to be in the region. In total, we manually analysed 820 of the 4929 (16.6%) files between 1 October 2018 and 24 April 2019 for the Ulukhaktok data. For Pearce Point, we analysed 486 of the 4682 (10.4%) files between 2 October (day the recorder was deployed) and 15 April. For Cape Bathurst at 50 m depth, we analysed 963 of the 4509 (21.4%) files between 10 October (day the recorder was deployed) and 15

Comparisons with other datasets
We compared trends in winter bowhead whale detections from multiple sites where we had previously collected data from the Amundsen Gulf. More specifically, we examined the final date in the autumn and the first date in the spring when a bowhead vocalization was detected. We previously collected and published results from three datasets: Sachs Harbour (71°55.621 0 N, 125°23.447 0 W) in 2015-2016 [30] and Ulukhaktok (70°42.857 0 N, 117°48.020 0 W) in 2016-2017 [31] and 2017-2018 [32]. All three of these acoustic datasets used the same type of acoustic recorder (SM3M) and the same mooring configuration and identical settings as the Ulukhaktok recorder described above, but with a recording duty cycle of one 5-min file recorded every 30 min and +18 dB of gain for the Sachs Harbour recorder.
We used the same automated detector and classifier on all of these datasets, as described above, and manually examined 100% of files with an automated detection for bowhead whales and a minimum of 10% of files without any automated detections in the autumn, winter and spring to determine when the final bowhead vocalization occurred during the autumn and first vocalization in the winter or spring occurred.

Sea ice concentration
We examined trends in winter sea ice concentration from 2013 to 2019 within a 100 km radius around the 2018-2019 acoustic recorder locations near Ulukhaktok, Pearce Point and the Cape Bathurst 50 m site using remotely sensed daily sea ice concentration data in 6.25 × 6.25 km grid cells from the AMSR-2 satellite array [42]. We examined the data between 15 November and 15 April of each winter in order to capture the majority of the time between ice formation and ice break-up every year. We calculated the average daily ice concentration across all grid cells within the 100 km radius around each site in each day. We then compared these values between years in order to quantify localized ice concentration.

Discussion
The evidence is clear that BCB bowheads overwintered in their summer foraging region in the eastern Beaufort Sea and Amundsen Gulf during the 2018-2019 winter and as far as we know, this is the first time it has been reported. Bowhead whales remained in the Ulukhaktok and Pearce Point areas for   [32,33]. Cape Bathurst would seem to be the more likely occasional overwintering location, being both a known summertime foraging congregation location for bowhead whales, particularly sub-adults, and a wintertime polynya [28]. Our recordings indicate that bowheads left the Cape Bathurst area in the fall, with a minimum of one month between October and November with no detections, but were detected there again every month of the winter  royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 8: 202268 between December and April. Whether or not the winter bowhead whales at Bathurst or Pearce Point were the same individuals as those who left Bathurst in October is unknown. There are several potential proximate drivers for the whales overwintering in the Amundsen Gulf. Ice entrapment has been documented with whales that were not ice-adapted [43], but is unlikely here as bowhead whales are known to move about in heavy ice conditions [16,30] and the east Amundsen Gulf had more ice-free water than normal during the winter of 2018-2019 (figure 4). Parasite and disease avoidance and molt have been raised as explanations driving whale migration [44,45] but seem less applicable for BCB bowheads whose migration is long (ca 6000 km) but largely east-west with relatively constant near-zero water temperatures. Predator avoidance is also worth considering. With warming ocean temperatures leading to decreased sea ice cover, killer whales (Orcinus orca) have been observed more regularly in the Chukchi and Alaskan Beaufort Seas [46] and are known to influence bowhead movement in the eastern Canadian Arctic [47]. Given the low levels of killer whale activity in the eastern Beaufort relative to the Bering and Chukchi Seas, killer whale presence could have been among the factors that led bowheads to remain in the eastern Beaufort Sea and Amundsen Gulf over winter in 2018-2019.
The potential of energetic savings from not migrating would need to be balanced with any change in food intake in order to result in a net benefit. Foraging that may have occurred during the fall migration   royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 8: 202268 or during the winter in the Bering Sea would need to be compensated with opportunities in the eastern Beaufort Sea, less the energetic cost of migrating. Bowhead preferred prey, such as copepods (Calanus spp.), are known to aggregate while hibernating near the sea floor and may have been available during winter if not too deep [19]. Overwintering would also ensure bowheads were present for the early summer plankton bloom, which may be limited in duration and occurring earlier and less predictably as the increasingly ice-free water allows greater light penetration into the water column [24]. In addition, if the overwintering whales were sub-adults of pre-reproductive age, as has been shown in other cetacean species [48], they would likely be smaller in size and thus may require less food intake than breeding adults. The relationship between water temperature and thermoregulation, independent of the water temperature-prey density relationship [7], could also be driving overwintering. Chambault et al. [49] found evidence for a narrow temperature range preference window between −0.5 and 2°C with the Eastern Canada-West Greenland stock of bowheads that was apparently more important than foraging and driven by the risk of hyperthermia when active (e.g. migrating) in warm water. Citta et al. [27] documented what may have been such an avoidance of the warm coastal waters associated with the Alaska Coastal Current by BCB bowheads. Anomalously high ocean temperatures also appear to be occurring with increased frequency. The summer following the overwintering observations described here, a widespread aerial census of bowhead whales in the Beaufort Sea conducted between July and October 2019 indicated unusually late westward migrating whales, many found further offshore than is normally the case [50]. It is not clear if this pattern of movement was attributable to warmer water temperatures or related to the 2018-2019 overwintering event. Such strong avoidance of warm waters have also been recently demonstrated in narwhals in the eastern Arctic [51,52]. If the avoidance of warm ocean temperatures were the primary driver of this anomalous behaviour, it may be a significant warning sign for bowhead whales. With this in mind, along with maintaining and expanding the locations of long-term recordings, our next step will be to correlate these data with ocean temperatures and other important oceanographic variables in the region.
Ethics. All data were collected under the authority of the Aurora Research Institute Scientific Research Licence No.