Temporal Variability of Sei Whale ( <scp> <i>Balaenoptera borealis</i> </scp> ) Acoustic Detections in Southern New England Waters

Sei whales (Balaenoptera borealis) are a large and endangered baleen whale species (Horwood 2009). They have a highly variable diet that can consist of copepods, euphausiids, crustaceans, and fish, with different prey preferences depending on the region (Horwood 2009; Mizroch et al. 1984; Prieto et al. 2012). Sei whales have a global distribution in temperate and subpolar waters, including along the United States (U.S.) Northeast coast (Prieto et al. 2012). Separate populations have been identified in the North Atlantic, North Pacific, and Southern Hemisphere (Pérez-Álvarez et al. 2021). The International Whaling Commission (IWC) recognizes two stocks of sei whales in the western North Atlantic for management purposes—the Nova Scotian stock (including the east coast of the United States) and the Iceland-Denmark Strait stock (Donovan 1991; Mitchell and Chapman 1977). Although the biological relevance of the stocks is inconclusive, and genetic evidence for the current division in the North Atlantic is lacking (Huijser et al. 2018), there is some evidence for at least two discrete feeding grounds: one off the Gulf of Maine and Nova Scotia and one in the Labrador Sea (Mitchell and Chapman 1977; Prieto et al. 2014). Sei whales undertake seasonal migrations, navigating from low-latitude wintering areas to high-latitude summer feeding grounds (Horwood 2009; Prieto et al. 2012, 2014). The whales that feed off the east coast of the United States and Canada have also demonstrated seasonal longitudinal movement across the North Atlantic (Olsen et al. 2009; Prieto et al. 2012, 2014). The locations of their wintering and calving grounds are unknown (Perry et al. 1999). One of the primary methods used to study sei whale presence and distribution has been through the use of passive acoustic monitoring (PAM), which is a continuous and cost-effective way to monitor the occurrence of vocal marine mammals (Zimmer 2011). Sei whales regularly make vocalizations, including a variety of tonal and broadband sounds (Baumgartner et al. 2008; Calderan et al. 2014; Cerchio and Weir 2022; Cusano et al. 2023; Mcdonald et al. 2005; Rankin and Barlow 2007; Tremblay et al. 2019). One call type in particular, the downsweep, has been consistently and definitively attributed to sei whales, allowing it to be used to detect sei whales with PAM in the North Atlantic (Baumgartner and Mussoline 2011). A downsweep is characterized by a continuous, descending frequency modulation from approximately 82 to 34 Hz, though the frequency range appears to vary depending on geographic location in the western North Atlantic (Baumgartner et al. 2008; Cusano et al. 2023; Macklin et al. 2024). Detections of sei whale downsweeps have been used to determine their seasonal presence and variability (Davis et al. 2020; Romagosa et al. 2020; Van Parijs et al. 2023), diel behavioral patterns (Baumgartner and Fratantoni 2008; Romagosa et al. 2020), and identify additional call types (Cerchio and Weir 2022; Tremblay et al. 2019). Recent passive acoustic studies revealed an increase in mean sei whale acoustic occurrence in southern New England feeding grounds in the years 2011–2014 compared to 2004–2010 (Davis et al. 2020), similar to the shift observed in North Atlantic right whales (Eubalaena glacialis) (Davis et al. 2017). The same study also demonstrated that sei whales were detected in southern New England waters year-round, with peak acoustic presence from March through July (Davis et al. 2020). Similarly, Van Parijs et al. (2023) found year-round acoustic presence in this area from PAM. The pervasive acoustic presence of sei whales in southern New England, and the increased use of the region by sei whales since 2010, identifies southern New England as an important area for sei whales that should continue to be monitored year-round. PAM is an effective tool for monitoring species, including sei whales, that have the potential for disturbance from offshore wind and other marine energy development (Van Parijs et al. 2023). As of January 2025, offshore wind energy is rapidly developing as a significant component of the clean energy transition along the coasts of the United States (Best and Halpin 2019; Snyder and Kaiser 2009). In the Northeastern United States there are currently 32 offshore wind energy projects in multiple stages of development between Maine and New Jersey (Musial et al. 2023). Despite the clean-energy benefits of offshore wind, impacts from construction and operation on marine ecosystems and their inhabitants, including marine mammals, must be monitored (Bailey et al. 2014). For example, activities such as pile-driving, drilling, and dredging are known to cause marine mammals to leave or avoid certain areas (Bergström et al. 2014). The increased vessel traffic and construction-related activities pose a heightened risk for noise exposure and vessel collisions, the latter of which are already a known threat to sei whales (Van Der Hoop et al. 2013). Monitoring for marine mammals in wind energy areas is crucial for assessing and mitigating potential risks and impacts on these animals (Bailey et al. 2014), particularly for endangered species like sei whales. Davis et al. (2020) and Van Parijs et al. (2023) combined passive acoustic data from multiple years to confirm year-round presence in southern New England. As sei whale presence is thought to be variable between years (Prieto et al. 2014), the goal of this study was to determine the seasonal presence of sei whales in this region using recent passive acoustic data and to compare across three consecutive years for seasonal and diel trends in sei whale downsweep detections. For this analysis, passive acoustic data were collected in southern New England waters from November 2020 through September 2023. Passive acoustic recorders (SoundTrap500, 600; Ocean Instruments Inc.) were deployed at three locations (COX01, COX02, and COX03) near Cox Ledge: south of Massachusetts and Rhode Island (Figure 1). The depths at the recording sites were relatively shallow (30–45 m). The recorders were spaced at least 15 km apart. As the detection range of sei whale vocalizations is estimated to be 10–15 km (Baumgartner et al. 2008), detections were unlikely to be heard on multiple recorders simultaneously; thus, calls were considered independent. All locations recorded continuously; however, the number of days of each deployment varied between 309 and 998 days (Table 1). The SoundTrap acoustic recorders exhibited a flat frequency response (±3 dB) between 20 Hz and 60 kHz and an effective recording range of 20 Hz to either 24 or 32 kHz depending on the sampling rate of the recorder (Van Parijs et al. 2023). All recorders had a system end-to-end calibration of −175.1 to −177.6 dB re 1 V/μPa, and self-noise that was < sea-state 0 at 100 Hz—2 kHz and < 36 dB re 1μPa above 2 kHz (Van Parijs et al. 2023). SoundTraps were attached 2–3 m above a fixed bottom mooring using VEMCO VR2AR acoustic receivers and weights, with subsurface floats extending ~6 m vertically into the water column (Van Parijs et al. 2023). For these deployments, all SoundTraps recorded at a sampling rate between 48 and 64 kHz (Table 1). The acoustic data were processed by the Low Frequency Detection and Classification System (LFDCS; (Baumgartner and Mussoline 2011)) to detect sei whale calls. The LFDCS downsamples the data (here, to 2 kHz), applies a low-pass anti-alias filter, and creates conditioned spectrograms using a short-time Fourier transform with a data frame of 512 samples and 75% overlap, resulting in a 64 ms time step and 3.9 Hz frequency resolution. The LFDCS then traces contour lines through tonal calls to create “pitch tracks,” each consisting of a time series of frequency-amplitude pairs. The pitch tracks are then classified into species-specific call types (based on a call library of known baleen whale calls in the western North Atlantic), using a multivariate discriminant function analysis (Baumgartner and Mussoline 2011). This study looked at sei whale downsweeps from the adapted GOM call library described in Davis et al. (2020). Each pitch track was assigned a Mahalanobis distance (MD) which quantifies the difference between a sound's pitch track and its assigned call type (Baumgartner and Mussoline 2011). A well-developed call type in an LFDCS call library will have 75% of pitch tracks with an MD of 3.0 or less (Baumgartner et al. 2013). Following Davis et al. (2020), an MD threshold of 3.0 was used for all sei whale downsweep detections to minimize the false detection rate, acknowledging that some true detections will be missed. All sei whale detections were manually reviewed by two acoustic analysts to limit the inclusion of low frequency downsweeps that may have been produced by other baleen whales (e.g., blue whales, Balaenoptera musculus, and humpback whales, Megaptera novaeangliae; Baumgartner et al. 2008; Berchok et al. 2006; Wilder et al. 2023). To avoid the potential misclassification of non-sei whale downsweeps, many PAM studies have made distinctions between “possibly correct” and “confirmed” detections. A downsweep that occurred as a single call (e.g., downsweeps greater than 5 s apart) was considered “possibly correct” although it falls within the frequency range of downsweeps produced by sei whales in this region (Baumgartner et al. 2008; Cusano et al. 2023; Macklin et al. 2024). A downsweep that was part of a doublet or a triplet (e.g., downsweeps separated by no more than 4 s from the start of one call to the start of the subsequent call) was considered a “confirmed” detection as doublets and triplets are considered unique to sei whales (Baumgartner et al. 2008; Davis et al. 2020; Van Parijs et al. 2023). To increase the certainty that sei whale presence was not underestimated in this analysis, we included any downsweep detection from singlets, doublets, or triplets, with each downsweep considered a single detection. Across all 3 years, there were 1840 validated downsweep detections. Sei whales were detected year-round in all 3 years where data were available (recordings ended mid-September 2023); though the seasonal peak presence varied by year (Figure 2). In 2021 and 2023, the highest detection rates occurred in spring (March–May), in 2022, the highest detections were in summer (June–August). Sei whales were consistently detected in low numbers in fall (September–November) and not detected at all in fall 2021 (Figure 2). Given the limited data collection at the end of the study, it is unclear if this trend continued into 2023, though even without a full year of data 2023 exhibited the highest number of detections and the highest rates of detected presence per week (Table 2, Figure 2). However, it is likely that the temporal distribution of sei whales in this region is highly dependent on additional variables other than season, particularly prey availability (Houghton et al. 2019; Payne et al. 1990; Schilling et al. 1992) as sei whales are known to aggregate in this region when feeding. Temperature also likely plays a role and is a strong predictor for sei whale presence in the North Atlantic (Houghton et al. 2019). Thus, peaks in seasonal presence will likely shift depending on these factors, as observed in North Atlantic right whales (Meyer-Gutbrod et al. 2023). Previous research found that sei whales in the Gulf of Maine were more vocal during the day when their preferred prey, Calanus finmarchicus, were at depth than at night when C. finmarchicus were near the surface where sei whales would likely be foraging (Baumgartner and Fratantoni 2008). Baumgartner and Fratantoni (2008) proposed that sei whale downsweeps are used during social behavior rather than during active foraging (Baumgartner and Fratantoni 2008), and downsweeps therefore likely indicate non-feeding behavior. For this analysis, the total number of calls per day was calculated separately for both day and night to assess diel trends. The light period designations were based on the sun altitude at each site using the suncalc package (Thieurmel and Elmarhraoui 2022) in R (R Core Team 2023). “Day” was defined as when the sun altitude was greater than or equal to zero and “night” was defined as when the sun altitude was below the horizon. The trends in calling rate were examined using a negative binomial generalized linear model (MASS R package; Venables and Ripley (2002)) to account for the notable amount of zeroes imposed by sei whale acoustic absence during the study period; both light period and month of the year were included as fixed effects (formula: Detections ~ Light Period * Month). Pairwise comparisons were conducted using the emmeans package (Lenth 2025) with the “mvt” adjustment. The data showed that sei whales in southern New England produced downsweeps both day and night throughout most of the year, with a notable decline in September and October evident in both light periods (Figure 3). Sei whales were often but not exclusively more vocally active during the day than at night (p = 0.850), with the overnight detected call rate typically only 0.70 times that of the daytime detected call rate (Table S1), although the call rates between light periods of the same month never exhibited a significant difference (Table S2). Despite the lack of statistical significance in call production between light periods, there are clear patterns evident in the data that suggest call detections are higher during the day than night in the summer (Figure 3), particularly in the summer of 2022 (Figure 2). Low sample sizes could have impacted the lack of statistical significance, and more data are needed to further parse apart trends. Previous research in the region identified reliable sei whale presence throughout the year, with considerable increases during spring and summer seasons (Davis et al. 2020). However, interannual fluctuations were not considered. Our study extends these observations by revealing seasonal fluctuations in sei whale peak presence, with potential aggregations in spring or summer depending on the year. Further, while our study revealed variability in call rates between day and night periods, it also demonstrated low detected call rates throughout the year, with 12 or fewer detections (about 1 downsweep detection per hour) in 86% of all light periods with sei whale acoustic presence. If sei whale downsweeps are a non-feeding call and sei whales primarily feed at night, there would likely be a statistically and biologically significant increase in downsweep production during the day; the consistently low call rates instead suggest the potential for feeding throughout a 24-h period. Such temporal variability warrants continued investigation as understanding the seasonal presence and behavior of sei whales is crucial for effective conservation efforts for this endangered species. Funding for the acoustic data collection and analysis from this study was provided by the NOAA through the Northeast Fisheries Science Center and Orsted Wind Power North America LLC. Additional helpful comments on the manuscript were provided by Julia Zeh, Andrea Jerabek, and Jason Selwyn. The authors declare no conflicts of interest. All acoustic data were collected with federal funding and are publicly available upon request. We are working to integrate the detection data into the publicly available Passive Acoustic Cetacean Map at https://apps-nefsc.fisheries.noaa.gov/pacm. Data S1. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.