Whistle Classification of Sympatric False Killer Whale Populations in Hawaiian Waters Yields Low Accuracy Rates

Abstract

Cetaceans are ecologically important marine predators, and designating individuals to distinct populations can be challenging. Passive acoustic monitoring provides an approach to classify cetaceans to populations using their vocalizations. In the Hawaiian Archipelago, three genetically distinct, sympatric false killer whale (Pseudorca crassidens) populations coexist: a broadly distributed pelagic population and two islandassociated populations, an endangered main Hawaiian Islands (MHI) population and a Northwestern Hawaiian Islands (NWHI) population. The mechanisms that sustain the genetic separation between these overlapping populations are unknown but previous studies suggest that the acoustic diversity between populations may correspond to genetic differences. Here, we investigated whether false killer whale whistles could be correctly classified to population based on their characteristics to serve as a method of identifying populations when genetic or photographic-identification data are unavailable. Acoustic data were collected during line-transect surveys using towed hydrophone arrays. We measured 50 time and frequency parameters from whistles in 16 false killer whale encounters identified to population and used those measures to train and test random forest classification models. Random forest models that included three populations correctly classified 42% of individual whistles overall and resulted in a low kappa coefficient, κ = 0.15, indicating low agreement between models, and the true population. Whistles from the MHI population showed the highest correct classification rate (52%) compared to pelagic and NWHI whistles (42 and 36%, respectively). Pairwise random forest models classifying pelagic and MHI whistles proved slightly more accurate (62% accuracy, κ = 0.24), though a similar pelagic-NWHI model did not (56% accuracy, κ = 0.12). Results suggest that the time-frequency whistle characteristics are not suitable to confidently classify encounters to a specific false killer whale population, although certain features of whistles produced by the endangered MHI population allow for overall higher classification accuracy. Inclusion of other vocalization types, such as echolocation clicks, and alternative whistle variables may improve correct classification success for these sympatric populations.

Skull morphological variation in a British stranded population of false killer whale (Pseudorca crassidens): a three-dimensional geometric morphometric approach

“…(Cranford 1999;MacLeod and MacLeod 2009;Ralls and Mesnick 2009;MacLeod 2010;Goldbogen et al 2019). A known trend is that the larger the animal, the louder sound it will produce (Ralls and (Murray et al 1998) and differences in vocalization were recorded between populations but not between sexes (Rendell et al 1999;Oswald et al 2003;Sanino and Fowle 2006;Barkley et al 2019). The allometric parallel slopes observed for males and females showed that the directions of shape changes in the cranium are conserved among sexes.…”

Dolphin communication during widespread systematic noise reduction-a natural experiment amid COVID-19 lockdowns

“…An analysis of Pearson's correlation coefficients revealed high linear correlation between delta frequency and high frequency (ρ = 0.94). Because previous studies have shown the inclusion of highly correlated variables impacts variable importance weightings in the model (Gregorutti et al, 2017;Barkley et al, 2019), we removed variables from the RFA if the Pearson's correlation coefficient exceeded ±0.8 as recommended by Barkley et al (2019). We used the Cohen's Kappa statistic (Landis & Koch 1977) to evaluate the model performance (Cohen, 1960) which measures the strength of agreement as follows; 0.01-0.20 slight, 0.21-0.40 fair, 0.41-0.60 moderate, 0.61-0.80 substantial, and 0.81-100 nearly perfect.…”

Tracking time differences of arrivals of multiple sound sources in the presence of clutter and missed detections

“…The amplitude part of the measurements for the targets and clutter is simulated by randomly drawing from the pdfs in Eqs. ( 11) and (10), respectively, with k ¼ 3:7 and the expected target SNR bounds ½d 1 ; d 2 ¼ ½5; 20 dB.…”

“…The clutter amplitude was simulated by randomly drawing from the pdf in Eq. (10), where k ¼ 3:7. The GM-PHD-SA tracking results showed that while the subgroup was tracked when only measurements from either whistles and (to some extent) clicks were available, better results were achieved when both measurements were combined (Fig.…”

“…1,5,6,59 False killer whales are highly vocal, producing broadband echolocation clicks 7,8 and narrowband whistles. 9,10 Due to significant differences in the signal properties, tracking and localization is typically carried out separately based either on whistles or echolocation clicks but generally not on both. However, as the types and rates of vocalizations depend on the animal behavioral context, 11 it is expected that a better understanding of a given acoustic encounter can be gained by performing the tracking task based on combined whistle and click information.…”