Establishing an acoustic benchmark of beluga habitat use in the Churchill River estuary, Manitoba

Belugas (Delphinapterus leucas) of the Western Hudson Bay (WHB) population migrate to the Churchill River estuary, Manitoba, Canada, forming the world’s largest summer aggregation. This estuary lies within an Ecologically and Biologically Significant Area and is proposed for inclusion in a National Marine Conservation Area (NMCA). However, the drivers of its persistent use by belugas remain poorly understood. Recent reopening and planned expansion of the Port of Churchill may increase vessel traffic and disturbance, underscoring the need to characterize beluga habitat use for impact assessments. This study presents the first passive acoustic monitoring (PAM) survey of beluga habitat use in the Churchill River estuary. Summer recordings were collected outside main port activity to establish a benchmark of beluga vocal behavior and habitat use. Vocal activity was greater during daytime than nighttime, varying with tide and river depth. Upstream activity increased during high tide, while downstream vocalizations were more frequent at low tide. These patterns were consistent with the use of shallow areas for social behavior and suggested movements toward deeper areas associated with socialization and feeding. Persistently high vocal activity in the deeper central estuary highlights its ecological importance and spatial overlap with port operations. Given the strong site fidelity of WHB belugas, this overlap raises concern about increasing beluga exposure to port activity and the potential for anthropogenic disturbance. A marked decline in vocalizations recorded around the time of ship activity may indicate that vessel traffic contributed to changes in beluga vocal activity, although concurrent weather conditions may also have influenced this pattern, hence warranting further investigation. Our findings confirm the Churchill River estuary as WHB beluga key summer habitat; demonstrate PAM as an effective beluga monitoring tool; and provide a critical pre-port expansion acoustic reference to support NMCA designation and beluga conservation.

Beluga whales (Delphinapterus leucas) are widely distributed across Arctic and sub-Arctic regions. Most populations are migratory, spending winters in partially ice-covered, often offshore waters and summers in river estuaries and sheltered coastal waters (Doan and Douglas, 1953; Caron and Smith, 1990; Smith et al., 2017; Shulezhko et al., 2018; COSEWIC, 2020). Of the 21 recognized beluga stocks, eight occupy or migrate to Canadian waters, including the Western Hudson Bay (WHB) population (Hobbs et al., 2019; COSEWIC, 2020). WHB belugas spend the winter in the Hudson Strait area and migrate to the river estuaries of WHB during the summer (Smith et al., 2017; Hobbs et al., 2019; COSEWIC, 2020) forming the world’s largest summer aggregation (Reeves et al., 2014).

The Churchill River estuary in Manitoba, Canada, is one of the key summer habitats for WHB belugas, who return annually in large numbers (Reeves et al., 2014; Pirotta et al., 2018). From June to early October, thousands of WHB belugas inhabit this area and nearby coastal regions, making the Churchill River estuary a well-known summer core area for this stock (Doan and Douglas, 1953; Sergeant, 1973; Idle, 1989; Matthews et al., 2016; Smith et al., 2017; Pirotta et al., 2018; COSEWIC, 2020). This contributed to the classification of the region as an Ecologically and Biologically Significant Area (EBSA) (Fisheries and Oceans Canada, 2011) and the recent inclusion of the estuary into a proposal for designation as part of a future National Marine Conservation Area (NMCA) (Government of Manitoba – Churchill NMCA). However, the reason for belugas’ annual occupation of river estuaries is not completely understood and may vary among stocks. Although feeding has been identified as the primary driver of beluga summer migrations in some river estuaries of the Russian Arctic (Litovka, 2002; Shulezhko et al., 2018), it does not appear to play the same dominant role in Canadian Arctic estuaries, including the Churchill River (Koski et al., 2002; Breton-Honeyman et al., 2016; Ausen et al., 2023).

Explanations for WHB beluga summer occupation of river estuaries include socializing, nursing, predator avoidance, molting, and juvenile development (Doan and Douglas, 1953; Caron and Smith, 1990; St Aubin et al., 1990; Watts et al., 1991; Krasnova et al., 2009; Chernetsky et al., 2011; Smith et al., 2017; Pitman et al., 2020; Whalen et al., 2020). To date, no single factor has been definitively identified as the sole driver of beluga occupancy in the Churchill River estuary, and the hypothesized drivers may not be mutually exclusive. Nevertheless, belugas show strong philopatry, returning to the same estuary each year and transferring knowledge of these summering areas to subsequent generations through learned behaviors (Smith et al., 1994; Colbeck et al., 2013; O’Corry-Crowe et al., 2020). This high level of site fidelity can make belugas susceptible to localized pressures, such as sustained harvest, that has contributed to depletion in some populations (Caron and Smith, 1990; Smith et al., 1994; Hammill et al., 2004).

Situated in the Churchill River estuary, the Port of Churchill reopened in 2019 after four years of inactivity during which only limited operational functions continued, such as sealift logistics (grainews.ca). As the only rail-serviced, deep-water port in North America with access to the Arctic Ocean, its maritime activities have recently expanded beyond grain shipping to include tourism, marked by the arrival of the first cruise ship in 2023 (CBC News, 2023). In addition, new development potential for the Port of Churchill has been recently announced, including upgrades to port infrastructures, new bulk-handling equipment, expansion of port facilities, dredging of the estuary, and potential ice-breaking operations to prolong the shipping season (Government of Manitoba). Given the pan-Arctic trend of increasing vessel traffic (Pizzolato et al., 2016; Dawson et al., 2018; Stevenson et al., 2019) coupled with the ongoing and proposed expansion of maritime activity in the Churchill River estuary, vessel traffic in this area is expected to rise rapidly. This will intensify beluga exposure to disturbance within this key summer habitat.

Vessel traffic is known to impact belugas in a multitude of ways, causing behavioral, acoustic, and physiological responses (Romano et al., 2004; Nowacek et al., 2007; Erbe et al., 2019). A major source of traffic disturbance is linked to the increase in underwater noise, which can interfere with beluga communication, navigation, socialization, and foraging (Finley et al., 1990; Lesage et al., 1999; Karlsen et al., 2002; Halliday et al., 2020; Vergara et al., 2021). Given the forecasted increase in vessel traffic at the Port of Churchill, improving our understanding of how belugas use this summer core area is critical. Establishing an acoustic benchmark of beluga vocalizations under limited disturbance conditions will allow for meaningful comparisons with future underwater recordings and will help identify potential changes in beluga vocal behavior and estuarine use in response to increasing anthropogenic activities and associated underwater noise.

In the absence of a documented acoustic benchmark for the area, this study aims to collect WHB beluga vocalizations in the Churchill River estuary during a period of low maritime operations, in order to derive temporal and spatial patterns of estuary use through passive acoustic monitoring (PAM). PAM was employed as a long-term, all-weather, and relatively non-invasive monitoring technique to record beluga underwater vocalizations during a period of reduced port activity associated with both the reopening of the port and travel restrictions related to the COVID 19 pandemic in summer 2020 (Ausen et al., 2022; Lasserre and Bartenstein, 2022). Under these conditions, whale-watching activity was limited to a small number of motorboats, and commercial ship traffic consisted of only a single vessel entering the estuary during the monitoring period.

The objectives of this study were to: (1) monitor beluga acoustic presence in two distinct areas of the Churchill River estuary during periods outside port main activity; (2) model beluga vocalization rates and investigate associations with environmental factors; and (3) infer potential beluga activities from their vocalizations. The results will contribute to conservation efforts for the WHB beluga population and to the sustainable development of the Port by informing vessel-traffic mitigation strategies in the Churchill River estuary.

Data collection

Acoustic data were collected in the Churchill River estuary, Manitoba, Canada, from August to September 2020 (Table 1) using single channel Porpoise hydrophones (Turbulent Research, Canada, https://turbulentresearch.com/). The hydrophones had a sensitivity of −200.83 dB re 1 V/µPa, recorded 24-bit WAV files, and were not calibrated. They were mounted on specifically designed moorings for shallow water and deployed from a small boat in two areas of the estuary – one in front of the Port of Churchill (H1) and one further upstream into the estuary, away from the main navigational channel (H2) – approximately 2 km apart (Figure 1). The moorings were designed to eliminate moving components that could generate noise and interfere with the detection of beluga vocalizations (Supplementary Figure S1).

Acoustic sample rate was set to 192 kHz to enable the recording of both beluga tonal sounds and clicks (Au et al., 1985; Sjare and Smith, 1986a; Chmelnitsky and Ferguson, 2012; Garland et al., 2015; Zahn et al., 2021). Although beluga echolocation clicks in the wild have been reported to contain spectral energy up to 120 kHz, hence exceeding the Nyquist frequency of the selected sample rate, most click energy was concentrated between 30 and 50 kHz (Au et al., 1985; Zahn et al., 2021), making the sample rate used in this study appropriate for reliable click detection. The recording duty cycle consisted of 29 minutes on and 1 minute off, which resulted in a total of 2720 files (1315 hours of recordings). Following preliminary in situ tests in the Churchill River estuary, the gain for both recorders was set to 15 dB. Further deployment details are presented in Table 1.

Water levels were extracted at hourly intervals from the Marine environmental data archive of the Government of Canada for the Churchill station (#5010, Institute of Ocean Sciences Data Archive, 2024). Because acoustic recordings were also collected hourly, water level and acoustic data could be accurately matched. To enable a comparison with other studies of beluga movements in estuaries, water level values were converted into tide phases (low, flood, high, and ebb tide) using a custom-built R code (R Core Team, 2024) implemented in RStudio (RStudio Team, 2024). The period that spanned ± 1 hour from the moment of the semidiurnal maximum water level was considered as high tide, while ± 1 hour from the minimum level was considered as low tide, with ebb and flood tides in between.

Daily weather data were downloaded from the Weather, climate and hazards database of the Government of Canada for the Churchill airport weather station (Climate ID: 5060595, Government of Canada, 2025), which is the closest meteorological station to the Churchill River estuary. Data of interest for this study were mean air temperature, total precipitation, and wind gust speed (defined in the Government of Canada dataset as maximum wind speeds exceeding 30 km/h).

Local light levels were extracted from the National Research Council of Canada’s sunrise/sunset calculator (National Research Council Canada, 2022). Day and night were delimited by local sunrise and sunset times, and twilight (civil twilight) was defined as the time of day when the center of the sun is six degrees below the horizon (National Research Council Canada, 2022).

Vessel presence was determined primarily through on-site visual observations. In addition, Automatic Identification System (AIS) data were investigated to determine the presence and location of AIS-equipped vessels in the Churchill River estuary during the study period. AIS data were extracted from the Arctic Ship Traffic Data (ASTD) database of the Protection of the Arctic Marine Environment (PAME) (https://pame.is/ourwork/arctic-shipping/current-shipping-projects/astd/). AIS data from the Hudson Bay Complex Large Marine Ecosystem area (PAME, 2013) were filtered spatially between the latitude 58.77° N and 58.80° N and the longitude 94.17° W and 94.24° W, and temporally from August 10 to September 11, 2020. AIS data were visualized in RStudio using base R functions.

Acoustic data analyses

For this study, beluga vocalizations were divided into two groups: tonal sounds and clicks (Au et al., 1985; Sjare and Smith, 1986a; Chmelnitsky and Ferguson, 2012). Vocalizations were identified using several automated detectors developed by JASCO Applied Sciences (https://www.jasco.com/). Six contour detectors were used to identify beluga tonal sounds: two for buzzes and four for whistles. Buzz detectors operated in high- (14–100 kHz) and low-frequency (1–10 kHz) bands with frequency resolutions of 64 and 16 Hz, respectively. Whistle detectors also operated in high- (4–20 kHz) and low-frequency (1–10 kHz) bands, with frequency resolutions of 64 and 8 Hz, respectively. Each whistle detector was run using two different detection thresholds, resulting in a total of six tonal sound detectors. Detection was based on a contour-following algorithm that identified tonal sound energy in the time–frequency domain and generated contours by linking adjacent bins that met specific criteria for each vocalization type (Delarue, 2025). Clicks were detected using an automated detector based on the Teager-Kaiser energy method, applied to high-pass-filtered (>5 kHz) raw acoustic data. Details for the click detector can be found in Frouin-Mouy et al. (2017) and Delarue (2025). Acoustic analyses were performed with the software PAMlab-Lite (JASCO Applied Sciences, https://www.jasco.com/pamlab).

Detector validation and performance

A subset comprising 5% of the acoustic recordings from each hydrophone location was manually analyzed to evaluate the performance of the automated detectors. Within each hydrophone dataset, files were randomly selected using a random integer number generator in R, and the same subset of files was used to validate both tonal sounds and clicks. No additional stratification by date, tide phase, or light level was applied. In total, 136 recording files were manually validated, and all analyses were performed by the first author to ensure consistency.

Recordings were examined both aurally by listening to the audio files, and visually by creating spectrograms using the sound analysis software Raven Pro, version 1.6.5 (Raven Pro, 2024). Spectrograms with 16384 samples per Fast Fourier Transform and a Hamming window with 75% overlap were created. This provided a frequency resolution corrected for windowing of 15.91 Hz and a temporal resolution of 0.02 s (Harris, 1978). Spectrogram contrast and brightness were adjusted during the analyses when required.

Beluga vocalizations were annotated by hand for comparisons with automated detections. For each recording, the first 30 s and 5 s were manually analyzed for tonal sound and click analysis respectively. These durations provided sufficient numbers of vocalizations for detector evaluation while keeping manual validation feasible. Given the high density of belugas in the Churchill River estuary during summer, recordings typically contained vocalizations from multiple individuals, and even short segments generally included large numbers of calls, particularly clicks, which often numbered in the hundreds.

Manual annotations were compared to the automated detections provided by JASCO by performing linear regression for the two groups of vocalizations separately. Detector performance was calculated by comparing the number of vocalizations detected by the algorithm with those identified by the first author. Vocalizations detected by both the automated detector and the manual annotation were classified as true positives (TP), those detected only by the automated detector as false positives (FP), and those identified only through manual annotation as false negatives (FN). Precision (P) and Recall (R) were then calculated following the formulas described by Sokolova and Lapalme (2009) and Hildebrand et al. (2022) to assess the performance of the automated detectors. Since different detector types were used for tonal sound and click identification, P and R were determined separately for the two vocalization groups.

Statistical analyses

Generalized linear mixed models (GLMMs) were used to investigate the potential correlation between beluga vocalizations and abiotic parameters, and run in RStudio using the package glmmTMB (Brooks et al., 2017; McGillycuddy et al., 2025). Separate models were run for clicks and tonal sounds. After preliminary analyses, a mixed linear-quadratic negative binomial distribution (nbinom12) (Brooks et al., 2017) was selected for both vocalization groups. The number of vocalizations detected per recording file was set as the response variable and tested as a function of tide phase and light level. These two variables were included as fixed effects, whereas hydrophone location was included as a random effect. Although only two hydrophone locations were included, hydrophone location was treated as a random effect to account for repeated measurements collected at each recording site and to control for location-specific variation unrelated to the fixed effects of interest.

Model optimization for both the fixed and random structure was conducted within each vocalization group. Model diagnostics were performed using the package DHARMA (Hartig, 2016). Models were run over a random subsample comprising 10% of the original dataset from each hydrophone location to reduce temporal autocorrelation while preserving the overall distributions of the response and explanatory variables. The 10% subsampling proportion was based on autocorrelation diagnostics, as it represented the smallest proportion that adequately reduced temporal autocorrelation. To verify that the subsampling procedure did not alter the distributions relevant to the analyses, the complete and subsampled datasets were compared using chi-square tests for tide phase, light level, and their joint distribution, and Kolmogorov–Smirnov (KS) tests for the distributions of click and tonal sound counts within each tide phase and light level group. In addition, descriptive statistics and histograms were compared between the complete and subsampled datasets. No statistically significant differences were detected, indicating that the subsample adequately represented the original dataset for the purposes of the GLMM analyses. However, as expected with random subsampling, rare high-intensity vocalization events were less likely to be retained and may therefore be underrepresented in the analyses.

To account for temporal autocorrelation and improve model fit, a first-order autoregressive AR(1) correlation structure was incorporated into the model for clicks. This factor was defined to correct for temporal dependence through date-time and operated independently within each hydrophone location. An ANOVA was run to compare the full and null models and test whether the explanatory variables significantly improved model fit.

The automated detectors performed well for both clicks and tonal sounds. Precision and recall were high for both vocalization type datasets (P = 0.85 and R = 0.91 for clicks, P = 0.79 and R = 0.91 for tonal sounds), indicating a high rate of correct detections and few missed vocalizations. Manual and automated detections were also strongly correlated (adjusted R 2 = 0.954 for clicks and 0.831 for tonal sounds; p<0.001 for both), with automated detections showing a slight tendency to overestimate vocalization counts relative to manual annotations (clicks: y=0.85x+2.83; tonal sounds: y=0.77x+2.67), (Figure 2). However, no bias correction was applied prior to the statistical analyses because detector performance showed high precision and recall, and the study focused on relative rather than absolute vocalization rates. Overall, these results support the reliability of the automated detections used in the subsequent analyses.

The number of clicks recorded at both hydrophone locations across the monitoring period was substantially higher than the number of tonal sounds, as expected for beluga acoustic behavior (Figure 3). Daily averages of tonal sounds followed similar general trends at both hydrophones (Figure 3, left), whereas click temporal patterns differed considerably (Figure 3, right). Click detections at H1 remained relatively constant throughout the monitoring period, whereas click detections at H2 decreased up to approximately two orders of magnitude beginning in mid-August. The average number of vocalizations detected per day exhibited a sharp decrease during the first week of September at both locations (Figure 3). Following this decrease, detections at H1 increased for both tonal sounds and clicks. Because the H2 mooring was retrieved during the same week as the vocalization decline, no data were available to determine whether a similar rebound occurred at that location. Consequently, late-season patterns after the early September vocalization decline can only be interpreted for H1.

To investigate potential drivers of the sudden beluga vocal decline, the sum of both clicks and tonal sounds recorded at each hydrophone were compared with local weather conditions and estuarine shipping activity (Figure 4). Mean air temperature showed a gradual seasonal decrease, going from 12.8 °C to 5 °C during the monitoring period. The highest precipitation event recorded (15.7 mm) was on August 30 th and 31 st, which also corresponded to the second highest strong wind event of the recording period, with gusts reaching 65 km/h. September 3 rd, marking the onset of the reduced vocalization event, coincided with a further rise in wind speeds that persisted throughout the week of beluga vocal absence. Wind intensity peaked on September 7 th at 80 km/h, the strongest wind event recorded during the monitoring period. That same day also experienced elevated precipitation (8.3 mm). Both high-wind events were associated with wind directions from the north and south, oriented parallel to the estuary (Supplementary Table 1).

Personal observations and AIS analyses indicated that two ships equipped with AIS systems were present in the Churchill River estuary during the study period. A harbor tugboat dedicated to Port of Churchill operations was present during the entire monitoring period (August 10 th to September 11 th, 2020). The vessel remained stationary (displacements < 5 m), except during the interval when a general cargo ship specialized in Arctic sealift operations (length: 137.16 m; Gross Register Tonnage (GRT): 8,448 MT) entered the estuary on the morning of September 8 th, 2020, and departed on the morning of September 9 th, 2020 (Figure 4).

GLMM analyses showed that light level and tide phase were both statistically significant predictors of beluga vocalization trends in the estuary. Nighttime corresponded to the lowest level of acoustic activity, with significantly lower values than at daytime for both tonal sounds (β= −0.8634 ± 0.1655 SE, z= −5.218, p<0.001) and clicks (β= −1.3303 ± 0.2672 SE, z= −4.979, p<0.001). Regarding tide phases, more beluga vocalizations were recorded during flood tide than during high tide for both tonal sounds (β= 0.8284 ± 0.2853 SE, z= 2.903, p<0.01) and clicks (β= 2.3217 ± 0.4624 SE, z= 5.021, p<0.001). Tonal sound detections during flood tide were also significantly higher than those recorded during ebb tide (β= 0.5012 ± 0.1422 SE, z= 3.525, p<0.001). Complete GLMM outputs are found in Supplementary Table 2.

Descriptive comparisons between hydrophone locations showed that both tonal sounds and clicks were consistently more numerous at H1 than at H2, regardless of time of day or tide phase (Figures 5, 6). Variability in beluga sounds was greater at H1 for both vocalization groups (Figure 5). At H1, beluga vocalizations decreased markedly from day to night, whereas day–night difference in both vocalization groups were less pronounced at H2 (Figure 5). Variation in vocal activity was greater across tide phases than across light levels (Figure 6). Summing all vocalizations, H1 exhibited maximum values during low tide, whereas H2 showed highest vocal activity during flood and high tide (Supplementary Figure S2). Tonal sounds peaked during flood tide at both hydrophones, with minima occurring during ebb tide at H1 and low tide at H2 (Figure 6, left). Click detections showed minimum median values during high tide for both locations, while maxima differed between low tide at H1 and flood tide at H2 (Figure 6, right).

Beluga tonal sound and click vocalizations recorded in this study provide new insight into habitat use of this species in the Churchill River estuary. The consistently high levels of vocal activity recorded throughout the monitoring season confirm the estuary as an important summer habitat for WHB belugas, supporting the designation of an EBSA that includes the estuary (Fisheries and Oceans Canada, 2011) and reinforcing the need to establish an NMCA (Manitoba Government). Persistent detection of both clicks and tonal sounds is consistent with the estuary supporting extensive social activity and communication, while opportunistic foraging is likely to occur primarily in the deeper downstream areas.

Both clicks and tonal sounds declined over the season, with a marked reduction in detections during the first week of September compared with the start of the monitoring period. This sharp decline potentially reflects the onset of beluga fall migration. Previous visual studies documented WHB belugas leaving the estuary between late August and early September (Doan and Douglas, 1953; Idle, 1989; Malcolm and Penner, 2011; Martin et al., 2001; Sergeant, 1973), consistent with our observations. More recent satellite-tagging work indicates departures occurring in early October (Pirotta et al., 2018), potentially linked to changing sea-ice conditions that prolong the open-water season (Andrews et al., 2017; Stroeve et al., 2024). Multi-year acoustic monitoring is needed to determine whether beluga migration timing in the estuary is shifting in response to climate change.

The sudden decrease in both clicks and tonal sounds during the first week of September coincided with the strongest winds and heaviest precipitation recorded during the study period. Temporary departures and reduced vocal activity during storms and high wind conditions have been documented elsewhere for belugas (Scharffenberg et al., 2019; 2020), while fewer individuals have historically been observed in the Churchill River estuary following periods of adverse weather (Doan and Douglas, 1953; Idle, 1989). A modest increase in vocalizations during the week following the event indicates that some belugas returned to the estuary, while others may have moved offshore. These observations suggest that deteriorating weather conditions may interact with the timing of seasonal migration and potentially act as a trigger for beluga departure. Climate-driven changes in sea ice phenology, particularly delayed freeze-up, have been linked to later autumn migration in belugas (see for example Hauser et al., 2017). However, climate change is projected to continue increasing both frequency and intensity of extreme weather events, including storms and wave events (Waseda et al., 2021; Wijngaard et al., 2025). These may conversely promote earlier beluga departures through mechanisms that may include strong winds and increased precipitation. Further research is needed to clarify the relative roles of sea ice dynamics and extreme weather events in shaping beluga migration timing, and to assess how their interaction may vary across stocks and regions.

Strong wind and precipitation may also have affected the detectability of beluga vocalizations by increasing ambient noise levels. However, no marked increase in background noise was apparent in the spectrograms during the period of decreased vocalizations, suggesting that reduced detectability alone is unlikely to explain the observed decline. Nevertheless, the relative contributions of weather-related changes in detectability and beluga behavior cannot be fully disentangled with the present dataset. Future studies combining visual observations with calibrated underwater noise measurements would enable a more definitive assessment of the relative contributions of beluga responses and changes in acoustic detectability.

Human disturbance, including vessel traffic, is also known to reduce beluga vocal activity and/or displace individuals (Halliday et al., 2019; Lesage et al., 1999; Vergara et al., 2025). During the monitoring period of this study, traffic from small motorboats remained relatively low and stable because of COVID-19 travel restrictions, and only a single cargo vessel entered the estuary following the reopening of the port. In addition to the strong weather conditions described above, the abrupt decline in beluga vocalizations observed in early September occurred shortly before the arrival of this vessel. Because belugas can detect vessel underwater noise over considerable distances (Cosens and Dueck, 1993; Erbe and Farmer, 2000), they may have perceived the approaching ship before it entered the estuary. Although vessel disturbance remains a plausible contributor to the marked decline in beluga vocalizations, its influence cannot be distinguished from that of concurrent weather conditions, highlighting the need for further studies to clarify the relative importance of these factors.

Our findings reveal diel variation in beluga vocal activity, with both clicks and tonal sounds significantly lower at night compared to daytime. Previous studies on belugas report various diel patterns, including daytime peaks (Booy et al., 2021; Castellote et al., 2013; Coppolaro et al., 2024; Giard et al., 2020) or no evident trends (Halliday et al., 2017). However, most lacked concurrent complementary surveys and therefore could not distinguish changes in vocal behavior from changes in animal presence. Only a few studies have coupled acoustic and visual monitoring to validate recorded vocalization trends. Among these, a study in Disenchantment Bay, Alaska, documented reduced nighttime echolocation despite constant beluga presence (Castellote et al., 2013), and similar decreases were observed in the Mackenzie Estuary (Scharffenberg et al., 2019). In our study, the absence of nighttime visual observations limits the interpretation of decreased nocturnal vocal activity, which may reflect beluga departure from the estuary or reduced vocal activity associated with resting or sleep (Lyamin et al., 2008; Panova et al., 2012; Castellote et al., 2013). Work in the St. Lawrence Estuary suggests that diel acoustic patterns may be primarily attributable to spatial shifts in beluga distribution across the estuary rather than intrinsic changes in vocal behavior (Giard et al., 2020). Because PAM alone cannot distinguish silence from absence, nighttime monitoring and satellite-tag studies would be valuable for validating diel variation in beluga presence and vocal behavior, hence supporting the development of a robust benchmark of diel activity.

Spatial variation in vocal activity between H1 and H2 reflected differences in bathymetry and tidal dynamics. Belugas are known to enter the Churchill River estuary during rising tides and exit during ebb tides (Doan and Douglas, 1953; Idle, 1989), a pattern that agrees with the distribution of vocalizations recorded in this study. H1, situated near the river mouth in deeper water, recorded the highest number of both vocalization types. Tonal sounds at H1 peaked during flood tide, suggesting increased use of the lower estuary as belugas entered the system; declined during high and ebb tides, consistent with belugas moving upstream; and increased again during low tide as belugas traveled back downstream. H2, located upstream in a shallower area, showed maximum tonal sound detections during flood and high tides and minimum detections at low tide, consistent with greater use of upstream areas when water levels were highest. Click detections followed similar spatial and tidal trends. At H1, clicks were most frequent at low tide and least frequent at high tide, with stable detections during flood and ebb tides. This pattern suggests that belugas may use the deeper downstream area as a movement corridor and take advantage of tidal flows to facilitate movement between estuary areas, relying on echolocation for navigation, as documented in other tidal estuaries (Ezer et al., 2008). At H2, click detections were higher at flood and low tide than during high tide, which may reflect increased echolocation requirements in shallow, turbid waters (Panova et al., 2012; Castellote et al., 2021).

Together, the acoustic detections indicate greater use of upstream, shallow areas during periods of maximum water levels and more persistent occupancy of downstream, deeper areas at low tide. Belugas have been documented using high tides to access shallow habitats that are unavailable at lower water levels. Previous studies show a preference for shallow waters (Castellote et al., 2013; Ezer et al., 2008) and upstream estuarine areas (Caron and Smith, 1990; Ezer et al., 2008; Kuznetsova et al., 2010; Shulezhko et al., 2018) during high tides, potentially to avoid stranding in shallow areas at low tide. The tidal movement patterns documented in this study are consistent with these findings. Furthermore, visual observations from earlier studies in the Churchill River estuary (Doan and Douglas, 1953; Idle, 1989) support the interpretation that the changes in vocalization rates recorded across tide phases and hydrophone locations primarily reflects shifts in beluga occupancy within the estuary rather than true diel patterns in vocal production.

Tonal sounds in belugas are typically associated with social interactions (Sjare and Smith, 1986a, b; Panova et al., 2012), whereas clicks are primarily linked to echolocation for navigation and prey detection (Au et al., 1985; Sjare and Smith, 1986b; Richardson et al., 1995; Rutenko and Vishnyakov, 2006). Maximum detections of both tonal sounds and clicks at H2 during flood tide suggest that upstream shallow areas may be associated primarily with social behavior, and navigation. This interpretation agrees with observations by Idle (1989), who proposed that upstream areas of the Churchill River estuary are important for molting. The high number of clicks recorded at H1 during low tide may indicate increased echolocation activity, potentially associated with navigation and/or opportunistic foraging, as tidal currents can promote prey aggregation within deeper channels (Simard et al., 2008). Although feeding might not be the primary driver of beluga estuary use in the study area, belugas have been observed foraging in deeper sections of the navigational channel and near the Churchill River mouth (Idle, 1989; Watts and Draper, 1986; Personal observations). However, because some studies have documented high click production in a variety of behavioral contexts, including social interactions (Sjare and Smith, 1986b; Panova et al., 2012; Castellote et al., 2021), future studies integrating PAM with direct behavioral observations, prey data, and improved characterization of beluga vocalizations, particularly buzzes, would help clarify the behavioral context of click detections and further evaluate these interpretations. The elevated occurrence of tonal sound vocalizations at H1 during low and flood tide suggests that the waters in front of the Port of Churchill are an important area for social interactions as well, further supporting the hypothesis that this part of the estuary serves multiple behavioral functions.

Collectively, these results demonstrate spatial variability in beluga habitat use within the estuary, with substantial overlap between areas of high vocal activity and port infrastructure. The reduction in human activity during the COVID-19 pandemic provided a unique opportunity to establish a critical acoustic benchmark of beluga vocal behavior in this key habitat, offering a valuable reference for informing vessel-traffic mitigation and management decisions on the potential expansion of the Port of Churchill and the recently proposed NMCA. Notably, the marked decline in vocalizations coinciding with the arrival of the cargo ship and associated movements of the tugboat suggests that vessel traffic may have contributed to changes in beluga vocal activity, although its influence cannot be separated from that of concurrent weather conditions with the present dataset. Together with recent evidence of increased use of the Churchill River estuary by other Arctic endemic species, such as the bowhead whale (Coppolaro et al., 2025), these findings have important conservation implications as maritime traffic in the region continues to increase, underscoring the need for precautionary, science-based management. Extending PAM in duration and spatial coverage, expanding vocalization classification, and integrating acoustic monitoring with prey data, behavioral observations, and telemetry data would refine our understanding of beluga movements and habitat use in this estuary, while enabling assessments of interannual patterns and potential changes in environmental drivers and responses to anthropogenic disturbance.

In conclusion, this study provides the first passive acoustic assessment of beluga use of the Churchill River estuary and confirms the area as an important summer habitat for WHB belugas, advancing our understanding of beluga habitat use in this region and contributing to broader knowledge of beluga estuarine ecology. Persistent detections of both echolocation clicks and tonal calls indicate that the estuary supports multiple important behavioral functions, including social interactions and potential opportunistic foraging. Beluga vocal presence was closely associated with tidal phase, light levels, and bathymetry, with shallow upstream areas used more frequently during high water and deeper, downstream areas during low water. These patterns agree with previous visual observations and highlight the effectiveness of PAM as a continuous, remote, and non-invasive approach for studying belugas in this important estuarine environment.

Consistently high vocal activity in deeper waters near the river mouth underscores the ecological importance of this part of the estuary and highlights its overlap with the Port of Churchill navigational channel within the EBSA. The acoustic baseline established in this study during a period of reduced vessel traffic associated with the early stages of port reopening and COVID-19 travel restrictions provides a valuable reference for assessing the effects of increasing maritime activity on beluga distribution and behavior in the Churchill River estuary. The marked decline in vocalizations observed close to the arrival of the first cargo ship, although confounded by deteriorating weather conditions, underscores the need for further investigation into this potential association. As maritime development expands in this coastal ecosystem, understanding how increasing human activities interact with natural environmental variability to influence beluga habitat use will be essential for effective management. Given the strong site fidelity of WHB belugas, evaluating the impacts of vessel underwater noise and growing maritime activities in the Churchill River estuary should be a priority to ensure that future port development remains compatible with the long-term conservation of this population and the ecological integrity of this key summer habitat.