AbstractHumpback whales (Megaptera novaeangliae) are a highly vocal baleen whale species with a diverse acoustic repertoire. “Song” has been well studied, while discrete “calls” have been described in a limited number of regions. We aimed to quantitatively describe calls from coastal Newfoundland, Canada, where foraging humpback whales aggregate during the summer. Recordings were made in July–August 2015 and 2016. Extracted calls were assigned to call types using aural/visual (AV) characteristics, and then agreement between quantitative acoustic parameters and qualitative call assignments was assessed using a supervised random forest (RF) analysis. The RF classified calls well (96% agreement) into three broad classes (high frequency (HF), low frequency (LF), pulsed (P)), but agreement for call types within classes was lower (LF: 63%; P: 85%; HF: 81%). We found support for a repertoire of 13 call types based on either high (≥70%) RF agreement (9 call types) or high (≥70%) AV agreement between two observers (4 call types). Five call types (swops, droplets, teepees, growls and whups) were qualitatively similar to call types from other regions. We propose that the variable classification agreement is reflective of the graded nature of humpback whale calls and present a gradation model to demonstrate the suggested continuum.
