Accounting for Sperm Whale Population Demographics in Density Estimation Using Passive Acoustic Monitoring PDF

Vol. 746: 121–140, 2024 MARINE ECOLOGY PROGRESS SERIES Published October 8 https://doi.org/10.3354/meps14693 Mar Ecol Prog Ser...

Vol. 746: 121–140, 2024 MARINE ECOLOGY PROGRESS SERIES Published October 8 https://doi.org/10.3354/meps14693 Mar Ecol Prog Ser OPEN ACCESS Accounting for sperm whale population demographics in density estimation using passive acoustic monitoring Alba Solsona-Berga1,*, Kaitlin E. Frasier1, Natalie Posdaljian1, Simone Baumann-Pickering1, Sean Wiggins1, Melissa Soldevilla2, Lance Garrison2, John A. Hildebrand1 1 Scripps Institution of Oceanography, University of California San Diego, La Jolla, CA 92037, USA 2 NOAA Southeast Fisheries Science Center, Miami, FL 33149, USA ABSTRACT: Sperm whales Physeter macrocephalus are highly sexually dimorphic, with adult males having larger bodies, more powerful echolocation clicks, and slower echolocation clicking rates compared to females. This study introduces methods for estimating sperm whale population densities in the Gulf of Mexico (GoMex) by accounting for the population demographics using passive acoustic monitoring and reveals that ignoring the differences between demographic seg- ments can introduce bias in density estimates. Weekly densities were estimated per 3 demographic segments: social groups consisting of adult females and their offspring, mid-size animals, and adult males. Analysis revealed that the GoMex sperm whale population is primarily composed of social groups, which account for 92 to 98 % of the overall population. Mid-size animals and adult males made up a small proportion of the population and were only intermittently present. Our 7 yr GoMex density estimates, including the 2010 Deepwater Horizon (DWH) oil spill period and sub- sequent years, revealed demographic-specific trends. Declines found at 2 north-central GoMex sites, coupled with increases at a southeastern site, may indicate population movements and poten- tial impacts from the 2010 DWH oil spill and elevated noise levels from anthropogenic activities in the north-central GoMex. KEY WORDS: Physeter macrocephalus · Sperm whale · Population demographics · Density estimation · Passive acoustic monitoring 1. INTRODUCTION 2019). Sperm whales in the northern Gulf of Mexico (GoMex) are managed as a separate stock under the Human pressures, including underwater noise, ship Marine Mammal Protection Act with an estimated collisions, fisheries interactions, pollution, and climate population of 1180 animals and an average density of change, hinder the recovery of sperm whale Physeter 1.7 animals 1000 km–2 (2017–2018 period; Garrison et macrocephalus populations (National Academies of al. 2020, Hayes et al. 2022). While still recovering from Sciences, Engineering, and Medicine 2017, Sousa et historical commercial whaling (Townsend 1935, al. 2019). Historical commercial whaling severely re- Reeves et al. 2011), this species faces contemporary duced global abundance by 57 % (Whitehead & Shin challenges related to habitat degradation. The GoMex 2022), leading to endangered status under the US En- region experiences exceptionally high levels of noise dangered Species Act, depleted under the US Marine pollution from seismic surveys for oil and gas explora- Mammal Protection Act, and Vulnerable under the tion and trafficked shipping lanes (Wiggins et al. 2016, IUCN Red List of Threatened Species (Taylor et al. Estabrook et al. 2016). Additionally, the lingering ef- © A.Solsona-Berga, K.E. Frasier, N. Posdaljian, S. Baumann-Pickering, S. Wiggins, J.A. Hildebrand and outside the USA the US Government 2024. *Corresponding author: [email protected] Open Access under Creative Commons by Attribution Licence. Use, distri- bution and reproduction are unrestricted. Authors and original publication must be credited. Publisher: Inter-Research · www.int-res.com 122 Mar Ecol Prog Ser 746: 121–140, 2024 fects of the 2010 Deepwater Horizon (DWH) oil spill — study (Solsona-Berga et al. 2022). This study revealed the largest in US history (Ramseur 2010, Levy & Gopa- spatial and temporal variability of the northern lakrishnan 2010) — pose a unique challenge to this GoMex population demographics, including sea- population. The consequences of these stressors re- sonal patterns and possible male presence year- main poorly understood (Farmer et al. 2018), high- round. lighting the need to monitor changes and inform man- Adult males, which have larger bodies, produce agement and conservation efforts. more powerful echolocation clicks at a slower rate Reliable estimates of population abundance or den- than adult females and juveniles (Goold & Jones 1995, sity trends are crucial indicators of species’ status in Solsona-Berga et al. 2022). Larger heads of adult the wild but remain challenging, particularly for pela- males may allow the buildup and discharge of greater gic species like sperm whales. Passive acoustic mon- volumes/pressures of air during sound production, itoring (PAM) is an attractive approach for remote resulting in higher source level emissions (Goold & long-term data collection, population estimation (Fra- Jones 1995). Differences in clicking rates between sier et al. 2016, von Benda-Beckmann et al. 2018), and sexes may be related to the maximum detection range trend analysis (Hildebrand et al. 2015). Assessment of for prey (Jensen et al. 2018). With a larger echoloca- animal density and abundance from autonomous tion range, adult males may slow down the interval PAM recorders relies on the detection and counting of between clicks to wait for more distant echoes. Differ- individual or group vocalizations in a given area and ences in clicking rates, detectability, and group sizes time period (Marques et al. 2013). Animal vocalization among sex/age groups (Douglas et al. 2005) may sig- rates, source levels, and group sizes are key scalar pa- nificantly impact acoustic density estimation of rameters required for accurate density estimation sperm whales. Calculating demographic-specific from such acoustic recorders (Marques et al. 2009, acoustic density estimates could reduce error intro- Hildebrand et al. 2015), and to date, these parameters duced by averaging parameters that are extremely are assumed to apply at the species or population different across demographic segments, and facilitate level. These assumptions may introduce bias in den- analyses of demographic-specific population trends. sity estimates, as sperm whales have complex popula- Using long-term PAM, we present a framework for tion demographics including social and sexual matur- estimating demographic-specific sperm whale den- ity segregation across different latitudinal ranges for sities using 2 approaches, cue and group counting, most populations (Best 1979, Lyrholm et al. 1999), as accounting for differing clicking rates, group sizes, well as sexual dimorphism in body size, which has and detectability among demographic segments. Our been linked to differences in echolocation click char- 7 yr GoMex density estimates, including the 2010 acteristics and diving behaviors (Gordon 1991, Wat- DWH oil spill period and subsequent years, revealed wood et al. 2006, Growcott et al. 2011, Solsona-Berga demographic-specific trends. Declines found at 2 et al. 2022). north-central GoMex sites and increases at a south- Adult females and immature animals form social eastern site could be linked to population move- groups in low and mid-latitudes (Best 1979, Rice ments, potential DWH oil spill impacts, and anthro- 1989). In contrast, maturing males form bachelor pogenic activities. groups of similar-aged animals and become increas- ingly solitary, moving to higher latitudes as they mature (Best 1979, Whitehead 2003). Males transit to 2. MATERIALS AND METHODS lower-latitude breeding grounds, but the timing re- mains poorly understood (Rice 1989). The GoMex Sperm whales were monitored at 3 northern GoMex population is mostly comprised of adult females and locations during and following the DWH oil spill immature animals, with smaller body and group sizes (2010–2017). These monitoring sites (Fig. 1) included than in other ocean basins (Jaquet & Gendron 2009). one near Mississippi Canyon (MC) within 15 km of Adult male movements and breeding times in this the DWH wellhead, another near Green Canyon regional population are still unknown. The little (GC), located outside and northwest of the DWH sur- knowledge we have, derived from studies of genetic face oil footprint, and a third near Dry Tortugas (DT), diversity and occasional visual observations showing outside and southeast of the oil footprint (Table S1 in both sexes moving across the GoMex basin, with the Supplement at www.int-res.com/articles/suppl/ some males breeding in different ocean basins (Lyr- m746p121_supp.pdf). holm et al. 1999, Alexander et al. 2016), is comple- At each site, a High-frequency Acoustic Recording mented by findings from a recent long-term PAM Package (Wiggins & Hildebrand 2007) recorded  Solsona-Berga et al.: Sperm whale demographics in density trends 123 Fig. 1. Acoustic monitoring sites in the northern Gulf of Mexico from 2010 to 2017, named for nearby oceanographic features: Green Canyon (GC), Mississippi Canyon (MC), and Dry Tortugas (DT). Deepwater Horizon wellhead is shown by a star, with cumulative surface oil footprint (Kobara 2019) in brown. Bathymetric contours at 1000 m depth increments are illustrated sound nearly continuously at a sampling rate of identified ship passages as times of increased noise 200 kHz. Sperm whale echolocation clicks were auto- in 3 specific frequency bands: 1–5, 5–10, and 10– matically detected and classified for use as cues for 50 kHz (see detector settings in https://github.com/ estimating weekly densities. All analyses, including MarineBioAcousticsRC/Triton/wiki/Ship-Detector). signal detection, classification, and subsequent den- Reduced recording effort resulting from the removal sity and trend analyses were carried out in MATLAB of these noise-dominated time periods was accounted R2016b (Mathworks). Signal detection and classifica- for in density estimates by adjusting recording effort tion was performed using an automated multi-step parameters to reflect the duration of data suitable for approach described by Solsona-Berga et al. (2022). sperm whale detection. Sperm whale detections were This approach implemented a band-pass filter be- grouped into encounters and manually validated in tween 5 and 95 kHz to reduce background noise. DetEdit software (Solsona-Berga et al. 2020) by 2 ana- Detections were filtered out at a peak-to-peak re- lysts (A.S.B., N.P.). While this process ensured classi- ceived sound pressure level (RL) of 135 dBpp re: 1 μPa fication accuracy of encounters, not every individual to establish a consistent detection threshold. Other detection was evaluated. Sperm whale social vocal- marine mammals like beaked whales or delphinids izations (codas, creaks, and slow clicks) and creaks for were automatically filtered out using spectral click foraging were not included in this analysis. The high characteristics such as peak frequency and spectral RL threshold of the detector prevented the detection shape (Solsona-Berga et al. 2022). Times during ship of codas and creaks, which have lower source levels passages were also excluded because sperm whale compared to regular echolocation clicks (usual clicks were often indistinguishable from ship-related clicks). Although slow clicks were detectable with noise. An automated vessel detector, Triton Ship- these settings, they were not found during the Det- Detector (Supplement of Solsona-Berga et al. 2020), Edit validation process. 124 Mar Ecol Prog Ser 746: 121–140, 2024 2.1. Population demographic classes Tkt represents the total time monitored at site k dur- ing week t, and rˆk is the estimated click production Sperm whale demographics were incorporated into rate at site k. density estimates to reduce potential sex/age class Group counting required detection of animal pres- bias by considering variability in clicking rates, de- ence within time windows (Hildebrand et al. 2015), tectability, and group sizes among size classes. Echo- along with knowledge of the detectability of clicks location click repetition rate, measured as the inter- produced by the observed group of animals. It relies click interval (ICI), served as a proxy for body length on knowledge of group size and group vocalization to infer demographic structure (Solsona-Berga et al. behavior. Using group counting, the estimated den- 2022, Posdaljian et al. 2024, Westell et al. 2024). This sity Dkt at site k, during week t is: method transformed ICIs into time series of modal ICI distributions within 5 min windows, associating dom- n kt `1–ctpkj`1 + ctnkj st Dt kt = r w 2 Ptk Ptv Tkt inant ICI patterns to body size classes. In this study, (2) we used the same time series of size classes found for GoMex sperm whales by Solsona-Berga et al. (2022). where nkt, cpk, cnk, and Pk are similar to those in Eq. (1) Sperm whale detections were categorized into 3 age/ but refer here in particular to time bins (5 min) with sex groups (referred to hereafter as demographic group detections rather than individual click detec- classes) in 5 min bins based on ICI distributions: (1) tions. ŝ represents the mean group size, and P̂ v is the small animals (