Global vessel traffic has increased rapidly in recent decades, driven by a growing merchant fleet and an increase in international maritime transportation, which now accounts for 80% of the total volume of trade goods (UNCTAD, 2022). As a result, underwater noise from vessel traffic has also intensified and is now nearly ubiquitous, reaching even marine areas that are far from major shipping lanes (Clark et al., 2009; Duarte et al., 2021; Thomsen and Popper, 2024). Once largely inaccessible, the Arctic is losing sea ice at unprecedented rates, facilitating a fast increase in vessel traffic in the region (Pizzolato et al., 2016; Andrews et al., 2018; Dawson et al., 2018; Stevenson et al., 2019). Maritime traffic in the Arctic has never been this high, nor its underwater soundscape so altered by human activities (PAME, 2024).

The main bandwidth of vessel underwater noise ranges from 10 Hz to 1 kHz (Richardson et al., 1995; Malakoff, 2010), overlapping with the frequency range used by many marine mammal species, which is typically between 10 Hz and 20 kHz (Erbe et al., 2018; Duarte et al., 2021). Recognized as a pervasive pollutant, underwater noise can negatively affect marine ecosystems and has been shown to cause a multitude of impacts on marine mammals, including communication masking, temporary behavioral changes, hearing loss, stranding, stress-induced health problems, permanent abandonment of biologically important areas, and, over time, potential consequences at the population level (Erbe et al., 2018; Southall, 2021; Tervo et al., 2021; Pirotta et al., 2022; Sweeney et al., 2022). Arctic marine mammals can be especially vulnerable to vessel noise (Moore et al., 2012; Hauser et al., 2018), making it critical to understand their behavioral responses to such disturbances.

The Southampton Island Area of Interest (AOI) in Hudson Bay, Nunavut, is under consideration for designation as a marine protected area (MPA) (Loewen et al., 2020). Vessel traffic in this region is increasing rapidly and is expected to grow further (Andrews et al., 2018; Dawson et al., 2018). The local community of Salliq (ᓴᓪᓕᖅ)/Salliit (ᓴᓪᓖᑦ) – Coral Harbour has expressed concern over the impact of vessels on marine mammals, particularly Atlantic walrus (Odobenus rosmarus (Linnaeus, 1758)) and beluga (Delphinapterus leucas (Pallas, 1776)) (COSEWIC, 2017; Carter et al., 2019; Loewen et al., 2020). The community has identified vessel activity as a major driver of changes in species distribution and abundance and called for changes in local maritime operations (Carter et al., 2019; Dawson et al., 2020). Gaining a better knowledge of vessel traffic impacts on marine mammals in this proposed MPA is essential for informed conservation efforts. However, current data on marine mammal habitat use in the AOI are limited (Loewen et al., 2020; Coppolaro et al., 2024), and no studies have yet investigated the acoustic responses of marine mammal species to vessels in the area.

Underwater passive acoustic monitoring (PAM) systems enable autonomous and continuous monitoring of marine mammals (Sousa-Lima et al., 2013; Heenehan et al., 2019; Halliday et al., 2020; Kline et al., 2020; Castellote et al., 2021) as well as the natural and anthropogenic sounds in their environment. PAM data can greatly contribute to assessing how marine species respond to vessel traffic. In Hudson Bay, studies examining walrus and beluga responses to vessels are limited and primarily based on visual observations (Caron and Smith, 1990; Mansfield and St. Aubin, 1991; Born et al., 1995; Malcolm and Penner, 2011; Ausen et al., 2022; Higdon et al., 2022). To date, no acoustic studies have examined walrus responses to vessel noise, and little is known about beluga vocal reactions to motorboats (Lesage et al., 1999; Karlsen et al., 2002).

This study aims to (1) integrate Automatic Identification System (AIS) data with PAM recordings to document vessel traffic in the AOI, distinguishing between ships and motorboats, and (2) investigate the underwater vocal responses of walruses and belugas to vessel transits in the Southampton Island AOI.

Walrus grunts and knocks, as well as beluga pulsed calls and whistles were detected during the study. Beluga were recorded in 8.1% and walrus in 5.9% of the total files. Vessel underwater noise was detected in 5.6% of the files. Beluga vocalizations were primarily recorded in June and November (Figure 2). Walrus vocal activity was low in June and increased in July, remaining relatively constant during the remainder of the monitoring period (Figure 2). Vessel noise was present throughout the entire study period, increasing throughout the summer and decreasing after October (Figure 2).

AIS data analysis revealed that 23 vessels entered the Southampton Island AOI during the monitoring period (Table 1, Figure 3). Most were general cargo ships and chemical tankers; no bulk carriers were detected (Table 1). Seven vessels, primarily general cargo ships, stopped in Salliq. Of the vessels transiting Evans Strait, 70% did not stop in Salliq but continued to other Hudson Bay communities, including Qamani’tuaq (Baker Lake), Igluligaarjuk (Chesterfield Inlet), Kangiqtiniq (Rankin Inlet), Tikirarjuaq (Whale cove), and Naujaat (Repulse Bay). A total of 50 vessel transits were detected in Evans Strait during the study period, 34 of which passed within the 10-km area around the hydrophone (Table 1). Ship traffic peaked between July and October, with only one or two trips detected in June and November (Figure 4). General cargo ships transited Evans Strait throughout the monitoring period, while chemical tankers were detected only between July and October (Figure 4).

Acoustic analysis detected 42 vessel transits during the entire monitoring period, of which 15 were classified as motorboats and 27 as ships based on their noise signature. When comparing AIS and PAM data, 15 of the 27 acoustically detected vessels classified as ships matched AIS vessel tracks that passed within the 10-km radius around the hydrophone. None of the vessels acoustically classified as motorboats corresponded with any AIS tracks. A total of 31 vessel encounters with walruses and 5 encounters with belugas were recorded (Table 2). No vessel encounters for either species were recorded in November and no beluga encounter with motorboats was detected for the entire monitoring period (Table 2).

Due to insufficient data, GLMMs could not be performed to assess the effect of vessel encounters on beluga call detection rates. As a result, modeling was limited to the walrus acoustic dataset. Among the models tested with different combinations of predictor variables, the model that included both noise exposure stage and vessel type, and the model with vessel type alone were equally supported as the best models based on AICc values (Table 3).

Walrus vocalization rates decreased during the transit of both vessel types, with significantly lower estimates during ship encounters compared to motorboat encounters (p<0.05) (Table 4). Walrus vocalization rates were highest before vessel transits and decreased during and after the encounters (p<0.1) (Figure 5, Table 4). No statistically significant difference was found between vocalization rates recorded during and after vessel encounters (Figure 5, Table 4).

Our results indicate that vessel traffic disrupted walrus vocal behavior. A reduction in walrus vocalization rates was found during encounters with both types of vessels, with ships causing more changes than motorboats. Previous studies indicate that walruses are vulnerable to vessel traffic (Moore et al., 2012; Erbe et al., 2018; Hauser et al., 2018), and that vessel noise overlaps with their hearing range in both air and water (Kastelein et al., 2002; Reichmuth et al., 2020; Duarte et al., 2021). Walrus hearing sensitivity is centered at frequencies between 1–12 kHz (Kastelein et al., 2002), which allows them to hear the main frequencies of both ships and motorboats in this study.

Most research on walrus vocal behavior has focused on the Pacific walrus (Odobenus rosmarus divergens), with documentation of their reactions to vessels being limited to visual observations of behavioral changes. The few studies on Atlantic walrus disturbance from vessels consist mainly of visual monitoring at haul-out sites (Born et al., 1995; Øren et al., 2018; DFO, 2019; Higdon et al., 2022). Reported reactions range from signs of short-term disturbance, such as head-raising and diving, to stampedes, shifts in feeding areas, and long-term abandonment of haul-out sites (Salter, 1979; Fay et al., 1984; Mansfield and St. Aubin, 1991; Born et al., 1995; COSEWIC, 2017; Higdon et al., 2022). To our knowledge, this is the first study investigating walrus acoustic responses to vessel transits.

Walrus populations subject to hunting are particularly susceptible to motorboat approaches (Malme et al., 1989; Higdon and Stewart, 2018; Øren et al., 2018). The community of Salliq relies on walrus for both subsistence and income (COSEWIC, 2017; Carter et al., 2019; Loewen et al., 2020), with motorboats commonly used for subsistence harvesting and sport hunting (COSEWIC, 2017; Minister of Justice, 2018). In this study, however, walrus vocalization rates were found to change significantly more during ship transits compared to motorboats, suggesting that motorboat traffic may have a lesser impact on walrus vocal behavior compared to ships. This difference may depend on the recurrence of exposure to the two vessel types, with motorboats being a long-established means of transportation in the study area, while consistent ship traffic is a more recent phenomenon (Dawson et al., 2018; Carter et al., 2019; Dawson et al., 2021). Repeated exposure to a stimulus can cause a decrease in the amplitude of marine mammal responses, a process known as habituation or acclimation (Groves and Thompson, 1970; Romero, 2004; Wright et al., 2007; Götz and Janik, 2011). The weaker response recorded for walruses during motorboat transits compared to ships may therefore reflect a process of associative learning from repeated events, as reported in other marine mammals (Born et al., 1995; Erbe et al., 2018; Harris et al., 2018). However, reduced behavioral reactions does not necessarily imply reduced stress and physiological responses (Wright et al., 2007), hence weaker responses should be carefully interpreted when developing disturbance mitigation measures.

This study could not determine whether the decrease in walrus vocalization rates during vessel encounters resulted from abandonment of the area or from a tendency to remain silent in those circumstances. Evans Strait is used by walruses for mating and foraging, while the southern coast of Southampton Island and the northern shore of Coats Island are important walrus haul-out sites (Carter et al., 2019; Loewen et al., 2020; Higdon et al., 2022; Coppolaro et al., 2024). Observed deviations from walrus undisturbed vocal behavior may be indicative of disturbance already occurring and should therefore be further investigated, especially in trafficked walrus hotspots such as the study area.

Future acoustic monitoring efforts could benefit from optimizing the recording duty cycle to align with the objectives of the survey and the acoustic behavior of the focal species. Specifically, longer listening durations or duty cycles with shorter intervals may enhance the detection of walrus and beluga vocalizations, improving assessments of acoustic presence, diel acoustic patterns, and call rate estimates (Thomisch et al., 2015). Such adjustments may also facilitate more accurate evaluations of marine mammal responses to anthropogenic underwater noise by better capturing potential changes in their vocal behavior. This would be particularly valuable in combination with effective assessments of noise exposure through measurements such as sound pressure levels. Furthermore, integrating PAM with AIS data and satellite telemetry from tagged animals would provide a more comprehensive understanding of vessel traffic dynamics and their impacts on marine mammals within the AOI (see, for example, Martin et al., 2024).

Belugas are notoriously a vocal species (Au et al., 1985; Sjare and Smith, 1986; Chmelnitsky and Ferguson, 2012; Panova et al., 2019). In this study, beluga vocalizations were detected more frequently than walrus sounds overall; however, detections were primarily concentrated during the months corresponding to beluga migration through the area (Carter et al., 2019; Loewen et al., 2020). In contrast, walrus vocalizations were spread across the entire monitoring period, reflecting the different habitat use of the two species. While the walrus is a resident species of the Southampton Island AOI, belugas mainly transit the area in early summer and autumn (Carter et al., 2019; Loewen et al., 2020; Coppolaro et al., 2024).

During this study, beluga vocalizations were never recorded during motorboat encounters and rarely during ship transits. Although most beluga vocalizations were recorded outside the main shipping season, AIS data and recordings of ship and motorboat noise confirmed temporal overlap between vessel transits and overall beluga acoustic activity in the study area. We hypothesize that the complete absence of beluga vocalizations during motorboat encounters may indicate an avoidance response, either physical or vocal. Previous studies have documented a decrease in beluga vocalizations in response to motorboat traffic (Lesage et al., 1999; Karlsen et al., 2002), as well as vocal interruptions in areas frequented by orcas (Orcinus orca) (Castellote et al., 2022). During motorboat transits, belugas may have reduced vocalizations to avoid detection by what could be perceived as a potential predator. Belugas were also known to leave an area when hunted from motorboats (Caron and Smith, 1990; Mymrin et al., 1999; Malcolm and Penner, 2011). Since belugas are subject to hunting in the AOI (Hoover et al., 2013; Carter et al., 2019; Loewen et al., 2020), the absence of beluga vocalizations during motorboat encounters in this study may also reflect their physical displacement. However, the data collected in this study were insufficient to assess beluga responses to vessel encounters. To investigate the aforementioned hypotheses, future research should include high-frequency analyses of beluga vocalizations—particularly ultrasonic burst pulses (Vergara et al., 2025)—which were beyond the scope of this work.

In this study, the combined use of PAM and AIS data allowed for the inclusion of both ship and motorboat information in the vessel traffic analysis. Including motorboats was deemed important for the purpose of this study due to both their common use and the growing presence of pleasure craft in the area (Dawson et al., 2018; Carter et al., 2019). Each monitoring methodology has its advantages and limitations. PAM allows for the detection of vessels without AIS instrumentation on board, which generally consists of motorboats, and of ships with AIS systems turned off (Kline et al., 2020). The acoustic analysis identified more vessel transits than those extracted from the AIS dataset, primarily due to the inclusion of motorboats exclusively in the PAM dataset. For ships, the detection discrepancy between the AIS and PAM datasets may be caused by the use of the exploratory 10-km hydrophone range. The selected range may have resulted in an underestimation of the number of AIS-tracked vessels that passed sufficiently close to the hydrophone to be acoustically detected. Also, AIS signal irregularities, equipment turned off, and the use of Class B units on some of the vessels may have contributed to inaccurate ship positions relative to the hydrophone (Corsi et al., 2023). Future studies would benefit from in situ measurements of the hydrophone sensitivity, local underwater sound propagation models, and from comparison tests between AIS and acoustic datasets using different distances for the hydrophone range (Aulanier et al., 2016). These would enable estimates of vessel and marine mammal distances, as well as a more accurate determination of the hydrophone detection radius to use for comparison with AIS data. Moreover, the use of multi-channel hydrophones would increase the number of synchronized listening points and enable localization of vocalizing individual positions in relation to vessels.

When applied to vessel traffic studies and compared to the use of AIS data, PAM limitations include a higher effort in extracting vessel position and speed, which often requires arrays or multichannel hydrophones, and the lack of certain navigation information such as vessel type (Zhu et al., 2018; Tesei et al., 2020). Manual analysis of recordings is time consuming, hence automated techniques are critical to significantly accelerate acoustic data analyses for vessel traffic studies (Reis et al., 2019; Vieira et al., 2020). The comparison between PAM and AIS data in this study also served to validate the classification method used to distinguish ships from motorboats based on their underwater acoustic signature. Overall, combining these two techniques demonstrated how passive acoustics can integrate standard marine traffic monitoring.

Vessel underwater noise was recorded throughout the entire monitoring period. Ship presence in early June and November indicates that local ice conditions allowed for marine traffic in the area, reflecting the prolonged ice-free season in the Arctic (Parkinson, 2014; Crawford et al., 2021). Interestingly, bulk carriers were not detected in the AIS data, despite the rapid growth of bulk carrier traffic in the Arctic (PAME, 2024). Most ships crossing Evans Strait in 2018 did not service the community of Salliq but transited the area to reach other Hudson Bay communities and the Baker Lake and Rankin Inlet mines. To reduce the impact of vessel traffic on marine mammals, vessels not servicing Salliq could be re-routed south of Coats Island, hence avoiding crossing biologically important areas. Agnico Eagle Mines Ltd. (2020) adopted such a rerouting plan for their sealift operations to the aforementioned mines, following the low-impact shipping corridors plan presented for the area by Dawson et al. (2018; 2020).

Disruption of individual activities can lead to long-term impacts on fitness and populations (Pirotta et al., 2018; 2022). As the ongoing reduction in sea ice is predicted to further boost Arctic marine traffic (Dawson et al., 2018; Stevenson et al., 2019; Rodríguez et al., 2024), additional studies are needed to better understand marine mammal reactions to vessel traffic in the Southampton Island AOI. For a marine protected area to be effective in supporting species health and conservation, measures must be implemented to minimize species disturbance (Williams et al., 2015). Vessel slow-downs and rerouting to avoid spatial and temporal overlap with key areas, especially during ecologically important periods, have been shown to reduce the impact of underwater noise pollution on several species (Pine et al., 2018; Williams et al., 2019; Findlay et al., 2023). Therefore, vessel traffic regulations and underwater noise mitigation strategies are deemed essential to support the establishment of a MPA around Southampton Island.