Interannual variability in acoustic detection of blue and fin whale calls in the Northeast Atlantic High Arctic between 2008 and 2018

Northern Hemisphere blue and fin whales are regular summer migrants to Arctic waters. Given the profound changes the Arctic is currently undergoing due to global warming, changes in habitat use and distribution of these migratory species are predicted. In this study, 3 passive acoustic recorders, 2 in Fram Strait about 95 km apart and 1 north of the Svalbard Archipelago (Atwain), were used to investigate the spatial and temporal vocal occurrence of these species in the Northeast Atlantic High Arctic. Acoustic data were available for 7 years for western Fram Strait (WFS), 2.5 years for central Fram Strait (CFS) and 3 years for Atwain. At both Fram Strait locations, most blue whale call detections occurred from August through October, though recently (2015−2018) in WFS a clear increase in blue whale call rates was detected in June/July, suggesting an expansion of the seasonal occurrence of blue whales. In WFS, fin whale calls were detected intermittently, at low levels, almost year-round. In CFS, fin whale calls were more frequent but occurred mainly from July through December. At Atwain, blue whale detections commenced in July, both species were recorded in September/October and fin whale calls extended into November. Results from this study provide novel long-term baseline information about the occurrence of blue and fin whales at extreme northerly locations, where traditional ship-based survey methods are seasonally limited. Continued sampling will support investigation of how environmental change influences cetacean distribution and habitat use.


INTRODUCTION
Blue whales Balaenoptera musculus and fin whales B. physalus have worldwide distributions ranging from tropical to polar regions (Cooke 2018a,b). These fast-swimming rorqual whales became the focus of commercial exploitation in the 1900s, due to the development of faster vessels and highly effective harpoons. During the modern whaling era, blue whales were hunted to the brink of extinction and fin whales were greatly reduced over their entire range. At present, Northern Hemisphere blue whale B. musculus musculus abundance estimates are still quite low in the North Atlantic region, apart from around Iceland and the Azores (Pike et al. 2009, Silva et al. 2013. A recent assessment of blue whales in the central North Atlantic estimated an abundance of 3000 (CV = 0.40) whales, suggesting an increasing trend in this area (Pike et al. 2019). Nevertheless, in the Northeast Atlantic, particularly around Svalbard and along the coast of northern Norway, blue whales continue to be rare and there has been no assessment of their numbers in this region since 1996 (Pike et al. 2009). Sightings of fin whales are more common in the North Atlantic and recent abundance estimates show clear evidence of recovery (e.g. Vikingsson et al. 2015, Pike et al. 2019. The North Atlantic population of fin whales is now likely close to, or larger than, the population was prior to the onset of modern whaling, with a total estimate of 79 000 individuals (CV = 0.13) in the North Atlantic (Cooke 2018b). While blue whales are protected, fin whales are harvested commercially at very low levels by Iceland (NAMMCO 2019a). Abundance estimates and distribution surveys for both species have focused on the region around the southeastern coast of Greenland and north of Iceland, with relatively little survey effort further north. However, there has been a notable increase in the number of reported sightings of migratory species, such as blue and fin whales, in the Svalbard area over recent decades (Storrie et al. 2018).
Both blue and fin whales are generally thought to be seasonal visitors to the Northeast Atlantic sector of the Arctic and are mainly observed during the open water season in summer/early autumn. Habitat use is not well known for either species in this area, but in general, these species are thought to arrive at feeding grounds at high latitudes (defined here as lattudes above the Arctic Circle, > 66°33' N) in late spring/ early summer and depart in autumn to spend the winters in more temperate areas . However, recent data from passive acoustic recorders suggest that some fin whales might overwinter at high latitudes (Simon et al. 2010, Ahonen et al. 2017. Sea ice is likely one of the main factors limiting the distribution of blue and fin whales, due to either physical obstruction of normal movement and breathing or secondary oceanographic effects on prey availability (Širović et al. , 2009(Širović et al. , Simon et al. 2010. However, as the Arctic is rapidly changing and the region is undergoing profound physical and biological changes associated with increasing air and water temperatures and decreasing sea ice extent, it is likely that habitat use and movement behaviour of the migratory whale species that frequent the Arctic seasonally will be affected.
Studying the geographic range and habitat use of a species is fundamental to understanding its ecology and predicting how it will respond to environmental changes (Gaston & Fuller 2009). However, this undertaking is especially challenging for highly mobile, pelagic species like blue and fin whales. In addition, vessel and aerial-based visual surveys in the Arctic are limited due to extreme seasonal variation in light conditions and inclement weather outside the narrow summer period. Visual surveys face challenges even in good light and weather conditions due to possible misidentification of species and responsive movement of animals towards or away from ap proaching vessels/aircraft (Pike et al. 2019). Tracking studies have been used at temperate latitudes to investigate distribution and habitat use in these wide-ranging cetaceans (e.g. Silva et al. 2013, Irvine et al. 2014, Jiménez López et al. 2019. However, the retention time of tags in these fast-swimming whales is relatively short, so animals tagged in temperate areas often do not retain tags long enough to reveal where they go when they come to the Arctic (e.g. Silva et al. 2013, Pérez-Jorge et al. 2020.
During recent years passive acoustic monitoring (PAM) has been used successfully to study the behaviour and distribution of many cetacean species in the Arctic (e.g. Moore et al. 2012, Clark et al. 2015, Ahonen et al. 2017, 2019, Hauser et al. 2017, Stafford et al. 2018. PAM allows for relatively low-cost, long-term monitoring throughout the year, under ice and in all weather conditions. PAM does have limitations, as call detection is influenced by the vocal behaviour of animals, and masking from natural and anthropogenic sounds influences what is detected (Stafford et al. 2007. Nevertheless, PAM provides the ability to collect year-round data on multiple species in otherwise inaccessible, remote locations (e.g. Stafford et al. 2007.
Over the past 2 decades, an increasing number of long-term PAM studies have been used to investigate distribution and migration patterns of blue and fin whales in various locations around the world, including polar regions (e.g. , Simon et al. 2010, Thomisch et al. 2016, Shabangu et al. 2019, Escajeda et al. 2020. Both blue and fin whales are good candidates for PAM because they produce readily identifiable stereotyped sounds. North Atlantic blue whales produce long (~20 s), low-frequency (16−18 Hz), frequency-modulated moans that can occur as repetitive 'songs' (Edds 1982, Mel linger & Clark 2003. Additionally, they produce higher frequency, shorter and more variable arch or D calls (Mellinger & Clark 2003). Blue whales vocalise regularly throughout the year, with peaks from midsummer into winter months (Stafford et al. 2001(Stafford et al. , Širović et al. 2004. Fin whales also produce low-frequency signals, but these are short (~1 s), frequency-modulated downsweeps that decrease in frequency from ~25 to 18 Hz and are often referred to in the literature as 20 Hz pulses (Watkins et al. 1987). These 20 Hz pulses can be produced either randomly or as repetitive bouts (singing) that can last hours to days (Watkins et al. 1987). Although fin whale vocalisations are detected year-round in some locations , Morano et al. 2012, Nieukirk et al. 2012, in the northwest Atlantic, fin whale singing has been recorded from June through March, peaking in October through December (Simon et al. 2010, Davis et al. 2020. The peak coincides with the fin whales' presumed reproductive season (Watkins et al. 1987(Watkins et al. , 2000. The purpose of this study was to investigate the acoustic presence of blue and fin whales in the Northeast Atlantic High Arctic using 2 PAM stations in the Fram Strait and 1 north of the Svalbard Archipelago (Atwain). Data from western Fram Strait (WFS) span a 10 yr period, allowing for investigation of possible temporal shifts in seasonal migration patterns for these 2 species. This novel long-term baseline data on occurrence of blue and fin whales in the High Arctic will be valuable for investigating how environmental changes are affecting the phenology of seasonally resident cetaceans.

Instrument type and deployment locations
Three Autonomous Underwater Recorders for Acoustic Listening (AURAL M2, Multi-Électronique Inc.; receiving sensitivity of −164 ± 1 dB re1 V μPa −1 with a 16 dB system gain and flat response from 5 Hz to 32.8 kHz, Kinda et al. 2013) were deployed on oceanographic moorings located in the Northeast Atlantic High Arctic (Fig. 1). Two of these moorings were situated in Fram Strait, 95 km apart. The westernmost mooring (hereafter referred to as WFS) is maintained annually by the Norwegian Polar Institute. It is situated at circa 78°50' N and 5°W, on the slope of the continental shelf (water column depth of ca. 1015 m). The AURAL was installed at a depth of 70−85 m (the exact depth changed slightly among years; Table 1). The central Fram Strait mooring (CFS) was maintained by the Alfred Wegener Institute and was situated at 78°50' N, 0°46' W (water column depth of ca. 2611 m). This AURAL was installed at a depth of 58 m on the mooring line. The third mooring was situated north of the Svalbard Archipelago at 81°30' N and 30°5' E (hereafter re ferred to as Atwain; water column depth of ca. 850 m). During the 2015− 2016 data collection period, this mooring was moved to shallower water (ca. 200 m, approximately 20 km away from original deployment location; see Table 1 for exact locations). This mooring is part of an instrumentation platform that provides data for the A-TWAIN project (https:// www.npolar.no/ en/ projects/ a-twain/), whose aim is studying long-term variability and trends in the Atlantic water inflow to the Arctic Ocean north of Svalbard. This AURAL was installed at a depth of 55−60 m on the mooring line (see Table 1 for exact depths for each location and year).

Acoustic data and detections
Acoustic data were available for 7 years for WFS (be tween 2008 and 2018), 2.5 years for CFS (between 2008 and 2012) and 3 years for Atwain (between 2012 and 2016). Data from WFS for 2008−09 were previously presented in Moore et al. (2012) but are reanalysed herein (i.e. the raw data were processed in a uniform way with the newer data) for comparative purposes. Sampling rate, duty cycle (number of minutes recorded at the start of each hour), and recording time varied among years and locations; full deployment details for each recorder are presented in Table 1. Because battery life is a limiting factor for these instruments, the longer duty cycle used during The spectrogram correlation modality in the program Ishmael (Mellinger & Clark 2000) was used as a pre-processing tool to detect times with low frequency signals for both blue and fin whales. Prior to analysis, all data collected with sampling rates greater than 8192 Hz were downsampled to 8192 Hz, and a low-pass filter (100 Hz) was applied. Additionally, a spectrogram equalization of 30 s was applied to reduce low-frequency noise from strumming and distant shipping. For blue whales, the spectrogram correlation parameters consisted of a 10 s long frequency modulated (FM) down-sweep from 18 to 16 Hz. The detection contour width was 3 Hz and a 1 s fast Fourier transform with 75% overlap was used. For fin whales, a 2 s long FM down-sweep from 30 to 19 Hz was used, with a contour width of 7 Hz. The resulting files with detections were then manually checked because all recording locations had some periods with low frequency noise. The data were corrected by removing all false positives and adding false nega-tives into the final dataset of blue and fin whale signals. Because the duty cycle varied among sites and years (Table 1), all blue whale de tections were normalized to call counts per hour for each day (standardised number of calls per day) by dividing the number of calls per day by total recording time (min) per day and extrapolating to the number of calls per hour (e.g. total calls per day/total minutes recording per day × 60). For fin whales, determining individual call counts was more difficult (individual calls were not as clear throughout the recording periods) so daily presence is reported as number of hours per day (0−24) with at least one clear fin whale call. For WFS and Atwain, the first 12 min of recording were used to count daily presence of fin whale calls (except for WFS 2008−09 which recorded 9 min each half hour and thus only the first 9 min period was used). For CFS, daily presence was assessed for the full recording time: 18 min for 2008−2009 and 14 min for 2010−2012. To ensure that no bias was introduced due to differences in the number of minutes used for each location, 25% of data from each year was reanalysed. Fin whale de tection results remained the same -at most only a few hours with detections each year were missed when sampling intervals were shorter. Based on these results, this is not considered to be a substantial issue. From the detection results described above, weekly averages were produced for both species at each location.

Temporal shifts in acoustic presence: model approach
For WFS, acoustic data were available for 7 years within a 10 yr period (from 2008 to 2018) allowing for investigation of possible temporal shifts in seasonal acoustic presence. Generalised additive models (GAMs) were fitted using the mgcv package (Wood 2011) in R (version 3.3.2; R Core Team 2016) to determine whether there were differences in call rates (calls per hour per day) between years. GAMs were selected as the best method to explore these data because they allow nonlinear relationships between predictor variables (in this case day, year, or period). Recordings started in September each year and lasted until the following August, except for 2012−13 and 2013−14 when recordings lasted only until April and in 2008−09 when recording continued until September 2009 (see Table 1). Temporal autocorrelation was examined using a first order autoregressive model (corAR1) structure in the model. As temporal autocorrelation was low (phi < 0.2), this term was not included in the final model selection.
Possible temporal shifts in acoustic presence were investigated by dividing the available data into 2 periods: early (data from 2008−2014) and more recent (data from 2015−2018). This division was considered to be the best way to split the 7 years of available data into 2 approximately equal data sets (4 data collection years in the early period and 3 in the latter period). A GAM was fitted with the response variable calls per hour (raw count, not standardised) and the predictor variables day and period. The variable day was included as a thin-plate spline smooth term and period as a 'by' factor variable within the smooth term to create separate smooth curves of day for each period (see Wood 2021 for more detail). Because the duty cycle varied between years (see Table 1), an offset term with variable time was included in the model. The distribution family used for this model was Tweedie (Tweedie 1984). Tweedie is a flexible distribution that can be used for different types of count data and is suitable for zero-inflated data (Wood et al. 2016). Tweedie behaves like a Poisson distribution if the P parameter equals 1, or can adopt a Gamma distribution when P equals 2. When using Tweedie in GAMs, the parameter P is auto-matically estimated by the model. The function predict.gam in the mgcv package was used to predict and compare call detections between the early and more recent periods.
To examine intra-annual patterns in acoustic presence, data for each year from WFS were divided into 2 periods: September to December (autumn/ early winter) and April to August (September for 2009; spring/summer). This division made it possible to exclude months with no detections (zeros in data set) given that instrument turn-around (changing batteries and redeploying the unit) usually occurred during September each year. A GAM with the response variable calls per hour (raw count, non-standardised) and the predictor variables day and year was fitted separately for these 2 time periods. The variable day was included as a thin-plate spline smooth term and year as a 'by' variable within the smooth term, so that a separate smooth term was made for each year. An offset term with variable time was again included in the model. The distribution family used for these models was Tweedie. After fitting the models, the function predict.gam was used to compare the call rates among years.

Distance to ice edge calculations
Because the presence of sea ice has been suggested as an environmental feature that excludes subarctic whales from the Arctic during seasons with heavy ice cover, daily sea ice extent images (Sea Ice Index data) were downloaded from NSIDC (Fetterer et al. 2017) and used to calculate minimum distance (km) from the mooring location to the ice edge (defined by the minimum 15% concentration contour, NSIDC). Daily GeoTIFF files for the Northern Hemisphere were downloaded for each recording period and processed in R using the rgdal and raster packages (Bivand et al. 2020 andHijmans 2020, respectively). Positive distances to the ice edge indicate the mooring was ice covered, whereas negative distances indicate the mooring was in open water.

Detection distance estimation
To obtain estimates of the range of distances over which blue and fin whale calls could be detected, the parabolic equation range-dependent acoustic model (Collins 1993) was used in Matlab R2012b (Math-Works Inc.) to model transmission loss in September. This month was selected as it was the month with the most consistent detections for both blue and fin whales. The purpose of this exercise was to provide maximum detection distances over which individually distinct calls could be detected, not conduct an indepth study of propagation loss in the study area. For blue whales, a frequency of 16 Hz and a source level of 179 dB re 1 μPa (Samaran et al. 2010) 1 was used. For fin whales a frequency of 20 Hz and a source level (SL) of 171 dB re 1 μPa (Charif et al. 2002) was used. It was assumed that both species vocalized at 30 m depth (Lewis et al. 2018). Minimum estimates of noise levels (NL) for each location were obtained: WFS 77 dB at 16 Hz and 79 dB at 20 Hz (Ahonen et al. 2017, this study), CFS 77 dB at 16 Hz and 78 at 20 Hz  and Atwain 76 dB at 16Hz and 74 dB at 20 Hz (this study). Because these values were all within 5 dB of each other, NLs of 77 dB at 16 Hz and 78 dB at 20 Hz were used. Sediment composition and sound speeds were obtained for Fram Strait from Hebbeln & Berner (1993). Critical (CR) values of 6 dB for blue whales and 10 dB for fin whales were set to achieve minimum signal-to-noise ratio for detectable calls. These CR values were estimated by selecting 15 blue and 15 fin whale calls from the data set that were just above the detection threshold and calculating signalto-noise ratios for these calls. Transmission loss (TL) plots from September were examined to determine the maximum distance over which calls could be detected based on noise levels at each location. For blue whales, the cutoff level was the distance at which propagation loss was less than 96 dB (SL -(NL + CR) = 179 -83) and for fin whales this was 83 dB (171 − 88). For a blue whale calling at 30 m depth, the maximum detection distance was ~100 km in deep water in mid-Fram Strait, but only 60 km near the shelf. For fin whales, these distances were 55 and 30 km, respectively.

Seasonal occurrence of blue whale calls
Across the study period, blue whale calls were detected in all sampling locations during the months of July, August, September and October (Fig. 2). Call detections varied across years in each location. Acoustic signals occurred in bouts that lasted hours to days. However, only at CFS (in 2010 and 2011) were blue whales vocal continuously over extended periods, i.e. nearly every day, from August through September (Fig. 3).

Western and central Fram Strait
Across all years, most blue whale call detections occurred from August through October at both Fram Strait locations (Figs. 2A,B & 3). Blue whale call detections were absent between November and May in CFS. In WFS, calls were absent between January and March and rare in November, December, April and May. In WFS, there was no clear pattern in weekly averages ( Fig. 2A). Overall, 2010 and 2011 had the fewest detections ( Fig. 2A). In 2016 and 2018, blue whale calls were detected more frequently as early as June, whereas in the earlier years for which summer data were available, frequent detections were mainly from August onwards ( Figs

Atwain
At Atwain, blue whale calls were recorded intermittently from July to October, with detections on 10 or fewer days each month (Fig. 3). Summer data (from June through August) were only available for 2016, whereas autumn data (from Septem- The source level for Antarctic blue whales was used in the absence of source levels for North Atlantic blue whales based on the characteristics of the signal and the relative size of the animals. We are assuming that similarly sized species might have similar sound production mechanisms. More importantly, both Atlantic and Antarctic blue whales make calls with similar characteristics -frequency modulated with songs composed of variations of a single note Weekly averages did not reveal any obvious pattern over the different years (Fig. 2C). Peak detections during autumn varied from mid-September to the beginning of October and during the year with summer data, the peak detection rate occurred in August.

Temporal shift in blue whale call occurrence in WFS
Seven years of data over a period of 10 yr from WFS al lowed for examination of potential temporal changes in blue whale call detections over a decade. There was a clear shift in blue whale call detections over time, with an increasing number of calls in more recent years (2015−2018) than in earlier years (2008− 2014), a seasonal shift toward calling earlier in the summer and a somewhat less prominent extension later into the autumn (Fig. 4 shows GAM results for the 2 periods; see also Fig. 3 for finer scale, monthly and annual patterns).
For the period September to December (autumn/ early winter) in 2008, the peak in call detection occurred towards the end of the first week and start of the second week of October (Fig. 5A). In 2012 and 2013, the timing of this peak was similar to 2008, but call rates were clearly higher than in 2008 and calling extended later into October. In autumn 2015 and 2016, call rate was higher in September than in October. In contrast, in autumn 2017, the peak in call rate occurred in mid-October, with calling ex tending into November. The number of calls per hour was clearly higher during the late October−early November period in 2017 than in other years.
For the period April to September (spring/ summer), there were no data for 2013 or 2014. How ever, in 2011 this period had clearly fewer call detections than the other 4 years and no clear peak call rate was ob served (Fig. 5B). The number of calls per hour in creased from the end of July through August and peaked in September in 2009. The call detection pattern in 2017 was fairly similar to that in 2009, but the number of calls was higher in August 2017 than in the same month in 2009. In 2016 and 2018, a very different pattern was observed than in other years, with call detections increasing in the beginning of June and being clearly higher than the other years until mid-August.

Seasonal occurrence of fin whale 20 Hz calls
Fin whale calls were detected seasonally in CFS and Atwain (summer/autumn), whereas in WFS they were detected intermittently throughout much of the year (Figs. 2 & 6). At all locations, individual loud fin whale signals occurred in bouts that lasted from hours to a few days.

Western and Central Fram Strait
Detections of fin whale 20 Hz calls varied between the 2 Fram Strait locations. At WFS, fin whale calls were detected throughout much of the year, though intermittently (not daily or weekly) and at very low levels (< 5 h d −1 ). No clear seasonal or temporal pattern was detected at this site ( Fig. 2A). At CFS, fin whale call detections had a clearer seasonal pattern and there were more hours per day with detections in peak periods (up to 23 h d −1 ). Most fin whale call de tections occurred from July through December, with peak occurrences in August and October (Fig. 2B). However, in January 2012 there were 11 days with detections (Fig. 6). Overall, the 2010−2012 recording period had more detections than 2008− 2009. Similar to the situation for blue whales, CFS and WFS are not directly comparable due to different recording dates. However, when both locations had data, CFS had many more fin whale call de tections and longer calling bouts than WFS (Figs. 2A,B & 6).

Atwain
At Atwain, fin whale calls were detected from the start of each sampling period (mid-September) through until early November (Fig. 2C). No calls were detected between December and July. Summer data (from June through August) was only available for 2016; however, no detections oc curred in June or July and only 2 days in August had fin whale calls (Fig. 6). Weekly averages did not re veal any consistent seasonal pattern over different years and peak detection times varied between years (Fig. 2C).

Temporal shift in fin whale call occurrence in WFS
Fin whale calls were detected intermittently through out most sampling periods in all years in WFS. However, no clear seasonal and/or temporal patterns in call detections were found (Fig. 6).

Distance to ice edge and detection of blue and fin whale calls
No clear patterns were seen regarding distance to the ice edge and occurrence of blue and fin whale call detections at any of the sampling sites (Figs. 3 & 6). However, ice edge plots showed that the ice edge was much closer to the mooring site at CFS than WFS (Fig. 3) and that CFS had more periods when the mooring was in open water (negative distance). Until recently, the WFS site has been ice-covered throughout the year; however, in 2017 between August and the beginning of October and in 2018 between the end of July and the end of August, the mooring was in open water. Distance to the ice edge varied considerably between months and years at Atwain during the data collection period. However overall, April seemed to be the month with the largest distance to ice edge.

DISCUSSION
This is the first multi-year study investigating acoustic presence of blue and fin whales in the Northeast Atlantic High Arctic. The PAM devices in this study sampled Fram Strait, the main gateway from the Atlantic Ocean to the Arctic Ocean, and a drift ice environment north of the Svalbard Archipelago above 81°N (Atwain). All 3 recording sites are in areas heavily impacted by global warming (e.g. Lind & Ingvaldsen 2012, Laidre et al. 2015. Detections of blue whale calls in Fram Strait between June and November have previously been reported , Haver et al. 2017). However, no previous studies included data from multiple years and, until the present effort, no PAM studies have been conducted further north than Fram Strait in the Northeast Atlantic Arctic. Visual surveys have been the only source of distribution and abundance information for these northern areas . In general, blue whale observations during visual ship-based surveys have been rare in this region (Øien 2009(Øien , Pike et al. 2009). However, a re cent study that included 13 yr of cetacean sighting data (collected between 2002 and 2014 from various citizen science sources; see Storrie et al. 2018) from around the Svalbard Archipelago suggested an increase in numbers of observations in recent years, although it is important to note that tourist operations also in creased in the region during the study period. But visual sightings are limited to months with enough daylight (March through November) and misidentification of species during visual surveys can lead to negative bias in blue whale observations and estimates (Pike et al. 2019). However, since blue whales are known to vocalise throughout the year and the identification of stereotypic low-frequency calls is relatively reliable, detected acoustic presence can be considered a good proxy of overall occurrence for this species. The estimates of the range of distances over which blue whales can be heard placed these animals a maximum of 100 km away (i.e. close to the mooring locations). Blue whales were detected seasonally at all 3 sampling locations, with most detections occurring in late summer/early autumn. Furthermore, acoustic data collected over a 10 yr period from the WFS site revealed an extended temporal presence of blue whales in recent years compared to earlier. GAM models showed that 2015−2018 had clearly higher numbers of blue whale calls per hour and that call rates in creased almost a month earlier (mid-June) than 2008− 2014. A temporal shift in autumn was not as clear; however, in 2017 calling bouts extended into November. Prior to 2017, blue whale calls were mainly detected only until mid-October and at considerably lower rates (Fig. 5A). Future studies are needed to verify this possible later departure of blue whales from Fram Strait. At all sampling locations, blue whales were absent from late autumn until early summer, so this species remains a seasonal migrant to the Northeast Atlantic High Arctic.
Similar to blue whales, fin whale calling has not been studied extensively in the Northeast Atlantic High Arctic. Fin whale chorusing (overlapping fin whale calls resulting in a continuous noise band) has been documented between September and March in central Fram Strait , Haver et al. 2017). These studies relied on a 'fin index' (Širović et al. , Nieukirk et al. 2012, which compares energy in the fin whale band to surrounding noise bands. However, this method detects chorus signals that can be hundreds to thousands of km away from the recorder. In addition, this method does not always distinguish between airgun signals and fin whale chorus signals. A previous study of the underwater soundscape in WFS reported fin whale 20 Hz chorusing between September and March (Ahonen et al. 2017). This study relied on the detection of fin whale signals within the 20 Hz energy band (and not individual 20 Hz calls). To be sure that fin whales were actually located close to the PAM sites, only loud, distinct individual calls were explored in the present study. This is likely why there was a lower daily presence and a shorter/different season reported herein than in Klinck et al. (2012) or Ahonen et al. (2017). The choice of method (chorus versus individual calls) is driven by the overall scientific questions being asked in different studies. Certainly, determining chorus is a much quicker, less labour-intensive method than detecting individual 20 Hz signals for fin whales. However, for this study, the interest was in the localized presence of fin whales (tens of km from the mooring versus many hundreds of km); therefore, the detection of individual pulses was selected.
Interestingly, the current study shows very different detection patterns for WFS and CFS. In WFS, individual fin whale calls were detected intermittently at low levels almost year-round, whereas in CFS, detections of fin whale calls were much more seasonal and more abundant (July through December). The WFS mooring was positioned within the East Greenland Current, which transports cold polar water southwards, while the CFS mooring was located in warmer Atlantic waters (Stafford et al. 2012). The western side of Fram Strait is seldom completely free of ice and is often covered with heavy sea ice, whereas ice conditions can vary greatly over short time periods at the CFS site (Stafford et al. 2012). However, ice conditions are generally lighter and distance to the ice edge is much shorter in CFS than in WFS (see for example Fig. 3). It is likely that fin whales prefer the open water conditions and warmer temperatures found in CFS, but this does not explain why fin whales are more seasonal in CFS. What determines their low-level interest year-round at WFS, despite the cold temperatures in this area, re mains unknown.
Ship-based visual surveys conducted in Norwegian Arctic waters have shown that fin whales are most often sighted west of Spitsbergen, along Eggakanten (the shelf edge), and are increasing in abundance (NAMMCO 2019b). Storrie et al. (2018) reported that fin whales were sighted around the Svalbard Archipelago and as far north as 81.5°N, but they have not been reported as far east as Atwain previously (Vacquié-Garcia et al. 2017, Storrie et al. 2018; Norwegian Polar Institute's Svalbard Marine Mammal Sighting Database, https:// data. npolar.no/ sighting/). In the present study, a seasonal pattern in acoustic detections of fin whales was seen at the Atwain site from September until mid-November. Fin whales seem to leave this High Arctic location a few weeks later than blue whales and also arrive later than blue whales. However, only 1 yr of summer data was available for Atwain so further information is needed to verify whether later arrival times are a trend.
Although fin whales were seasonally detected in CFS and Atwain, the fact that individual calls were detected throughout much of the year in WFS confirms that some fin whales are staying at high latitudes throughout the year. This is also supported by results from a recent, novel tracking study in Svalbard, where 40% of tagged animals did not initiate southwesterly migrations during their tracking periods (September−November; Lydersen et al. 2020). In Davis Strait, fin whales are acoustically present until January (Simon et al. 2010) and there was a negative correlation between fin whale calling and sea ice cover, suggesting that they migrated when sea ice formed. In the current study, no clear pattern emerged between distance to ice edge and fin whale call detections, but the increasingly open, drift-ice nature of the ice cover in Fram Strait likely facilitates better conditions for these whales, which are not ice-adapted in the same manner as the Arctic endemic cetaceans.
Sea ice loss in the Arctic is one of the most visible climate-induced changes on the globe (Notz & Stroeve 2016). With the loss of sea ice, there is potential for spatial and temporal increases in habitat for seasonally migrating temperate cetaceans (Moore 2018, Stafford 2019. The Barents/Greenland Sea region is an Arctic 'hot spot', experiencing some of the most rapid declines in the seasonal extent of sea ice in the Arctic, driven in part by higher atmospheric and ocean temperatures (Kelly et al. 2010, Pavlov et al. 2013, Nordli et al. 2014, Onarheim et al. 2014, Laidre et al. 2015. A recent study by Storrie et al. (2018) found signs of northward range expansions in seasonally resident species such as blue and fin whales in Svalbard waters. These findings follow a broader pattern of range expansions of marine animals that has been described as Atlantication or borealization of the North Atlantic/Barents Sea region, associated with increasing water temperatures and decreasing sea ice across the North Atlantic (e.g. MacLeod et al. 2005, MacLeod 2009, Lambert et al. 2014, Fossheim et al. 2015, Vikingsson et al. 2015, Vihtakari et al. 2018, Polyakov et al. 2020. Spatial range expansion could not be investigated with the acoustic data from this study; however, GAM results for WFS suggest a possible temporal/seasonal expansion in blue whale occurrence. The most parsimonious explanation for this is that sea ice loss and warmer waters are making the region more suitable habitat for a longer period in the summer and fall and also causing shifts in prey abundance and availability . In these times of rapid ecosystem alteration, baleen whales such as blue and fin whales can serve as sentinels of climate change impacts through changes in their phenology, distribution, body condition and abundance (Moore 2018). Passive acoustic sampling is one of the recommended methods to collect data from highly mobile, vocal species such as baleen whales, particularly from remote locations such as High Arctic areas .
This study provides novel information on acoustic presence of temperate cetaceans in High Arctic waters, information that would be difficult to obtain with traditional survey methods. The Atwain site represents the northernmost acoustic recording of blue and fin whales to date. Additionally, this study shows that blue whales seem to have increased their use of high latitudes by arriving earlier in the year and possibly departing later. Furthermore, this study confirms that at least some fin whales are staying at high latitudes throughout the year, demonstrating the utility of PAM to collect important phenological and distributional information that would otherwise remain unknown.