Sperm whales exhibit sexual dimorphism and sex-specific latitudinal segregation. Females and their young form social groups and are usually found in temperate and tropical latitudes, while males forage at higher latitudes. Historical whaling data and rare sightings of social groups in high latitude regions of the North Pacific, such as the Gulf of Alaska (GOA) and Bering Sea/Aleutian Islands (BSAI), suggest a more nuanced distribution than previously understood. Sperm whales are the most sighted and recorded cetacean in marine mammal surveys in these regions but capturing their demographic composition and habitat use has proven challenging. This study detects sperm whale presence using passive acoustic data from seven sites in the GOA and BSAI from 2010 to 2019. Differences in click characteristics between males and females (i.e., inter-click and inter-pulse interval) was used as a proxy for animal size/sex to derive time series of animal detections. Generalized additive models with generalized estimation equations demonstrate how spatiotemporal patterns differ between the sexes. Social groups were present at all recording sites with the largest relative proportion at two seamount sites in the GOA and an island site in the BSAI. We found that the seasonal patterns of presence varied for the sexes and between the sites. Male presence was highest in the summer and lowest in the winter, conversely, social group peak presence was in the winter for the BSAI and in the spring for the GOA region, with the lowest presence in the summer months. This study demonstrates that social groups are not restricted to lower latitudes and capture their present-day habitat use in the North Pacific. It highlights that sperm whale distribution is more complex than accounted for in management protocol and underscores the need for improved understanding of sperm whale demographic composition to better understand the impacts of increasing anthropogenic threats, particularly climate change.

Passive acoustic recordings were collected at seven sites, two along the BSAI and five in the GOA, between June 2010 and September 2019 (Figure 1; Table 1). Each site had from one to ten deployments which resulted in ~12 years of cumulative recordings between all sites. Individual site temporal coverage varied due to project goals, recorder battery life, data storage space, and duty cycle regimes as detailed below. The sites were in moderate water depths of 780 m to 1200 m (Table 1). We used High-frequency Acoustic Recording Packages (HARPs; Wiggins & Hildebrand 2007) with a sampling rate of 200 kHz which can detect the high-frequency echolocation clicks of odontocetes, including but not limited to, sperm whales.

Sperm whale echolocation clicks were detected using the multi-step approach described in Solsona-Berga et al. 2020 (appendix). These clicks have multiple pulses (Backus & Schevill 1966), 2-9 ms apart, depending upon the size of the animal (Norris & Harvey 1972). As a result, the detector had a lockout for clicks separated by less than 30 ms to avoid multiple detections of a single click. Band passing the data (5-95 kHz) minimized the effects of low-frequency noise from vessels, weather, or instrument self-noise on detections, but allowed for detection of the echolocation clicks of toothed whales. The Power Spectral Density (PSD) of detected signals was calculated with the Pwelch method (MATLAB, MathWorks Inc. 2016) using 4 ms of the waveform and a 512-point Hann window with 50% overlap (Welch 1967). Instrument specific full-system transfer functions were applied to account for the hydrophone sensor response, signal conditioning electronics, and analog-to-digital conversion. To provide a consistent detection threshold, only clicks exceeding peak-to-peak (pp) sound pressure level (RL) of at least 125 dBpp re 1 µPa were analyzed. This threshold was chosen to eliminate noise signals and the echolocation clicks of other odontocetes, while retaining sperm whale clicks.

Sperm whale echolocation clicks can be confused with the impulsive signals from ship propeller cavitation. An automated classifier developed by Solsona-Berga et al. 2020 (appendix) was used to exclude periods of ship passages. The classifier identified potential ship passages from long-term spectral averages (LTSA), which are long duration spectrograms (Wiggins & Hildebrand 2007). Further averaging was calculated as Average Power Spectral Densities (APSD) per 2-hour blocks over low (1-5 kHz), medium (5-10 kHz), and high (10-50 kHz) frequency bands with 100 Hz bins and 50% overlap. Using received sound levels, transient ship passage signals were separated from odontocete echolocation clicks and weather events. A trained analyst manually reviewed identified ship passages using the MATLAB-based custom software program Triton (Wiggins & Hildebrand 2007). Ship passage times were removed from further analysis and considered time periods with no effort.

Instrument self-noise and the echolocation clicks of other odontocetes were also removed to reduce the number of false positive detections. A classifier using spectral click shape was implemented, taking advantage of a sperm whale click’s distinct low-frequency spectral shape to remove dissimilar clicks by delphinid and beaked whales, which typically have higher frequencies (Solsona-Berga et al. 2020). The remaining acoustic encounters containing putative sperm whale echolocation clicks were manually reviewed with DetEdit, a custom, MATLAB-based graphical user interface (GUI) software program used to view, evaluate, and edit automatic detections (Solsona-Berga et al. 2020).  

Histograms of ICI provide a visualization that can be used to indicate sperm whale size and sex (Solsona-Berga et al. 2022). A plot of concatenated histograms, referred to as ICIgrams, was annotated and categorized for each time period at each site. Examples of the ICIgram GUI can be found in Solsona-Berga et al. (2022). We used three ICI groups to correspond to three size classes (Fig. 2, bottom panels), as per Solsona-Berga et al. (2022). Detections with a modal ICI of 0.6 s or less were presumed to be females and their young, hereinafter referred to as Social Groups. Detections with a modal ICI of 0.8 s and greater were presumed to be adult males, hereinafter referred to as Adult Males. The detections with a modal ICI between the Social Groups and Adult Males (< 0.6 s and > 0.8 s) could contain large females or juvenile males, hereinafter referred to as Mid-Size.

Sperm whale click detections were binned into 5-minute intervals. The mean daily presence per week was calculated by summing the number of 5-minute bins with detections for each size class and for each site. Since not all sperm whale clicks were categorized into a size class, a time series of unclassified clicks was also included for each site. The ratio of hourly as well as daily presence for each size class was calculated and displayed with Venn diagrams to show the overlap of the classes at each site. Finally, these data were grouped into one-hour bins for statistical modeling, as described in the next section. The one-hour bins were chosen as a compromise to maintain data granularity while ensuring at least 30 minutes of recording effort in each one-hour bin for the two duty-cycled deployments.