Song structure and singing activity of two separate humpback whales populations wintering off the coast of Caño Island in Costa Rica

Humpback whales (Megaptera novaeangliae) migrate seasonally between feeding and breeding habitats (Clapham, 2018). There are currently 14 distinct population segments (DPSs) recognized worldwide (Bettridge et al., 2015). Of these, two migrate to the Pacific Central American waters to breed, the Central American population (DPS 6) and the Southeastern Pacific population (DPS 13), also known as Breeding Stock G (International Whaling Commission, 1998). The Central American (CA) population migrates from feeding areas in the Northern Hemisphere, specifically California and Oregon, to breeding areas off the Pacific coast of Central America during the boreal winter (∼December–April) (e.g., Steiger et al., 1991; Calambokidis et al., 2000; Rasmussen et al., 2012). In 2016, the National Marine Fisheries Service (NMFS) declared this population as threatened (NMFS, 2016). Whales from the Breeding Stock-G (BSG) are seen off Costa Rica and Panama in the austral winter (∼June–September) and migrate from feeding areas in the Southern Hemisphere, specifically off the Antarctic Peninsula and the Fuegian Archipelago in Chile. The population migrating from Antarctica to Costa Rica represents one of the longest known migrations of any mammal (Acevedo-Gutierrez and Smultea, 1995; Rasmussen et al., 2007; Acevedo et al., 2017).

On the South Pacific coast of Costa Rica, the waters of the Osa Península, and in particular the north and east sides of Caño Island are considered important breeding habitats for both CA and BSG humpback whales (Steiger et al., 1991; Rasmussen et al., 2007; Palacios-Alfaro et al., 2012). Photo-identification data (photos of unique natural markings on the flukes of humpback whales) suggest that most CA humpback whales arrive in this area in January and depart in April (e.g., Calambokidis et al., 2000). BSG whales were first reported at Caño Island and Osa Peninsula during the months of August and September (Acevedo-Gutierrez and Smultea, 1995; Palacios-Alfaro et al., 2012; Rasmussen et al., 2007). Although both populations are genetically distinct (Jackson et al., 2014), molecular data provides evidence for gene flow (Medrano-Gonzalez et al., 2001) albeit infrequent and limited to the edge of their wintering seasons (Jackson et al., 2014). However, Rasmussen (2006) reports year-round sightings of humpback whales along the Pacific coast of Costa Rica raising the prospect of overlap between these populations in this region.

Humpback whales are highly vocal animals that produce a variety of social sounds with songs being the most studied and prevalent type of acoustic behavior produced at breeding areas (Dunlop et al., 2007). These songs are long and complex and only produced by males (Payne and McVay, 1971). Within a distinct population, males typically conform to the song of their conspecifics over a breeding season (Herman, 2017). The songs display high geographical and seasonal stratification, with song sharing only occurring between regions that share individuals (Noad et al., 2000; Garland et al., 2011). Song learning can take place even when the number of individuals involved in the exchange is low (Garland et al., 2011).

This study provides the first documentation of daily and monthly humpback whale' song detection and description of song structure for both CA and BSG humpback whales wintering off Caño Island in the South Pacific coast of Costa Rica. Given the potential for temporal overlap, this study also explores the possibility of horizontal transmission of song elements between these populations. Understanding the acoustic behavior of these whales at the edge of their wintering areas can contribute to the study of song transmission and population connectivity in the Pacific Ocean.

The study took place within the protected waters of Caño Island Biological Reserve located 15 km off the Osa Peninsula in the South Pacific coast of Costa Rica. Recorders were deployed in two locations, Jardin and El Diablo at a depth of 25 m. Jardin (8.719 N/-83.863W) is 0.8 km from the island on the northeast side and is characterized by a sandy bottom. El Diablo (8.70 N/-83.915W) is 2.15 km from the island on the north side and is characterized by a rocky reef at the edge of the reserve. The distance between these sites is approximately 6 km. Recordings were made in a 24-h duty cycle in both locations from three to 11 days. At the Jardin location recordings were made from September 19–30, October 1–14, 24–31, November 1–5, December 7–19, 2016 and January 12–31, February 7–18, March 21–31, April 20–30, June 19–27, 2017. At El Diablo recordings were made from January 12–22, February 7–17, March 21–31, April 1–9, 20–30, May 1–3, June 1–3, 27–30, and July 1–9, 2017. The recorders used were the SM2M+ (Sampling rate: 4–96 kHz −165 dB re 1 V/μPa) from Wildlife Acoustics (2019) and the RUDAR-mK2 (Sampling rate up to 96 kHz −169 dB re 1 V/μPa) from Cetacean Research Technology (1994). The recorders were deployed interchangeably to reduce field costs and maximize recording time. Recorders were programmed to continuously capture the soundscape at a sampling rate of 44–48 kHz in 30-min files (except for September due to an error in programming, recording files were 5-min long) for a total recording effort of 2670.5 h at Jardin and 1771.2 h at Diablo. To determine whale-singing activity, a 1-min sample was manually taken every five minutes for a total of 30 186 1-min files for Jardin and 21 520 files for Diablo. These files were uploaded to ARBIMON II (Sieve Analytics, 2015) for cataloguing and spectral inspection for presence/absence of whale songs. The proportion of song presence was calculated by dividing the number of 1-min files with songs by the total number of 1-min files per day and month for each whale population.

Humpback whale song structure for both populations were analyzed using one full song cycle from recording files with a single whale. All selected songs from the BSG had a signal-to-noise ratio (SNR) >8 dB while the CA whale songs the SNR ranged between 4 and 6 dB. While the SNR for CA was not optimal, we included their songs in the analysis if they had recognizable song components and harmonic content (Mercado et al., 2003). Songs were inspected independently by two observers (EC and LJMC) using spectrograms in Raven 1.5 (2016; Cornell Lab of Ornithology) with a fast Fourier transform size of 2048 points, an overlap of 50%, and a 512-sample Hann window. Upon agreement between observers, whale song elements were classified into units, phrases, and themes following standard descriptions of humpback whale song components (e.g., Cholewiak et al., 2013). Song elements were labelled sequentially using roman numbers for themes, capital letters for phrases and Arabic numbers for units. Phrases and unit labelling are distinct for each population. The following acoustic variables were extracted for each unit: minimum, maximum, and peak frequency (Hz) and duration (s) (e.g., Perazio et al., 2018). The statistical software JMP 14.2.0, 2018 (sas Institute, Inc.) was used for descriptive statistics and to plot the proportion of recordings with song presence using a smoothing spline and bootstrap confidences of fit with a lambda of 0.05.

A total of 665 h of acoustic time were analyzed during the breeding season of CA humpback whales from December 2016 to May 2017. Only 35 1-min files contained whale songs (n = 25 in December 2016 and n = 10 in March 2017), representing <1% of the recording time analyzed. Thirty-three of these detections occurred between 02:00 and 05:00 a.m. and two at 23:00 p.m. For BSG whales a total of 206 h of acoustic time were analyzed for June and July 2017, which resulted in no song detections. From September to November 2016 a total of 196 h of acoustic time was analyzed resulting in 150.3 h containing whale songs (76%). The proportion of BSG whale song detections varied with month: 34% in September, 78% in October and 18% in November [Fig. 1(a)]. BSG whale songs were detected throughout the day, but diel patterns varied between months [X² = 29.1, df = 2, p < 0.0001, Fig. 1(b)] due to significant differences in the proportion of song detections in September between midnight and midday (p < 0.0001).

A total of 64 full song cycles (n = 21 September, n = 32 October, n = 11 November) were analyzed corresponding to 16 song sessions. Overall, BSG whale song encompassed 4 themes, 10 phrases, and 12 units (Fig. 2, Table 1). Together, themes I and III made 75% of the song in September, 64% in October, and 69% in November (Table 1). Song structure changed slightly during the study period with two phrases gained (C and J) by mid-October (Table 1). Once C was acquired, it became an important component of the song (October = 23% and November = 35%) (Table 1). Overall, song units had a mean minimum frequency of 617.3 ± 439.3 Hz, mean maximum of 2940 ± 1302.4 Hz, mean peak frequency of 1684.1 ± 756.6 Hz, and mean duration was 1.85 ± 1.29 s. Table 2 summarizes the acoustic characteristics of each unit.

A total of five full song cycles were analyzed for CA whales, two from December 2016 and three from March 2017. The CA whales' song consisted of 2 themes, 3 phrases, and 4 units. Song structure changed within the season, in December the song consisted exclusively of theme I, which consisted of a single phrase, K1. In March, the presence of K1 decreased to 34% and a similar phrase K2 was added to theme I. A second theme was detected in songs from March and consisted of a single phrase, L, which made 37% of the song (Fig. 2). Overall, song units had a mean minimum frequency of 261.2 ± 243.6 Hz, mean maximum of 421.2 ± 289.7 Hz, mean peak frequency of 33.3 ± 261.7 Hz, and a mean duration was 2.52 ± 1.64 s. Table 2 summarizes the acoustic characteristics of each unit.

The study finds acoustic evidence of BSG whales remaining in the study area in November and documents the first occurrence of CA whale's acoustic presence in December. While in general the song structure of each whale population is distinct, both populations shared the second part of unit 7 (phrase E) of BSG whales and unit 16 (phrase L) of CA whales. In addition, some variations of unit 1 from BSG whales look similar to unit 13 from CA whale songs (Fig. 2).

Our results find that while both CA and BSG humpback whales are acoustically present in Caño Island during their respective breeding seasons, song detection is higher for BSG than for CA whales. This is consistent with their reported time of arrival and occurrence in the study area (Steiger et al., 1991; Calambokidis et al., 2000; Rasmussen et al., 2007; Rasmussen et al., 2012; Palacios-Alfaro et al., 2012). Although the study did not cover the first two months of the BSG whales breeding season (July and August), our results show that song detection peaked in October and not before. Similar patterns of song occurrence have been reported in Albrohols, Brazil (Sousa-Lima et al., 2018). Regarding daily song detection, BSG whales sang throughout the day, but with a decrease between midnight and midday hours. Similar singing patterns have been reported in other breeding grounds (e.g., Au et al., 2006).

Humpback whale song's theme diversity varies by region and with time. For example, songs from whales recorded in Hawaiian, Mexican, and Panamanian breeding areas consisted of six themes (e.g., Cerchio et al., 2001; Oviedo et al., 2008), while those recorded in Colombia consisted of three themes (Perazio et al., 2018). Songs recorded in other locations within the Southern Pacific Ocean basin ranged between four and eight themes (Garland et al., 2013). The BSG and CA whales recorded in this study are at the lower end of theme diversity with four and two themes, respectively. However, CA theme diversity is likely underestimated due to the small sample size and recording quality. As described for other populations (e.g., Perazio et al., 2018) one theme dominated most of the song and structural changes were found in song composition including themes being sang outside the standard theme order, addition of phrases, and large standard variation values for unit frequency and temporal characteristics, all consistent with seasonal and individual variation in song structure.

Humpback whales are sighted year-round in the Pacific coast of Costa Rica (Rasmussen, 2006), raising questions about temporal overlap between CA and BSG whales (Rasmussen et al., 2007). Based on song structure, this study finds evidence for late departures of BSG whales in November and early arrival of CA whale songs in December, setting the time of potential overlap sometime in mid-late November. The late departure of BSG whales is also supported by unpublished photo-identification data of a whale photographed the last week of November at Caño Island by JDPA that has matched the Antarctica Peninsula catalog (Happywhale, 2015) (photo of individual HW-MN1300572 by V. Beaver), and early arrivals of CA whales is supported by whales photographed in early December that matched the California catalog (Calambokidis, 2019). Unfortunately, we could not re-deploy the recorders during the last two weeks of November 2016 due to weather conditions, and a recorder deployed from June 2018 to June 2019 was lost before the date of recovery. Despite the potential for acoustic connectivity between these populations, this study finds no strong supporting evidence for this hypothesis. The song of each whale population was distinct, and while two units were shared, sharing two units is not definitive evidence for cross-equatorial acoustic connectivity as these units may represent broadly shared templates. Interestingly, no similarities were found in song structure with two other BSG whale's songs studies from Colombia (Perazio et al., 2018) and Panama (Oviedo et al., 2008). Comparing published spectrograms which figures vary in resolution is not optimal and future efforts should be made for raw data comparisons.

This study provides the first description of diel song detection and song structure of BSG and CA whales wintering off the South Pacific coast of Costa Rica. The results contribute to ongoing efforts to establish the conservation status of both populations in Central America. Despite acoustic and photo-ID evidence for late departure of BSG and early arrival of CA whales during the month of November, this study finds no evidence for cross-equatorial acoustic connectivity. However, it warrants further exploration.

This paper is dedicated to Gretchen H. Steiger in gratitude for her pioneering work on the Central American humpback whale population in the Osa Peninsula, Costa Rica and for inspiring so many scientists in the region. We are forever thankful, rest in peace dear Gretchen. Funding came from Conservation International. Thanks to Marco Quesada Alpizar, Juan Jose Alvarado, and Sebastian Mena González for field and logistic support; Ingi Agnarsson, Jeff K Jacobsen, and Kristin Rasmussen for manuscript feedback; John Calambokidis, Ted Cheeseman, and Vicki Beaver for photoID information; and to two anonymous reviewers for comments that greatly improved this manuscript. This study was done under permit No. ACOSA 033-16 MINAE.