Categorizing click trains to increase taxonomic precision in echolocation click loggers.

Passive acoustic monitoring is an efficient way to study acoustically active animals but species identification remains a major challenge. C-PODs are popular logging devices that automatically detect odontocete echolocation clicks. However, the accompanying analysis software does not distinguish between delphinid species. Click train features logged by C-PODs were compared to frequency spectra from adjacently deployed continuous recorders. A generalized additive model was then used to categorize C-POD click trains into three groups: broadband click trains, produced by bottlenose dolphin (Tursiops truncatus) or common dolphin (Delphinus delphis), frequency-banded click trains, produced by Risso's (Grampus griseus) or white beaked dolphins (Lagenorhynchus albirostris), and unknown click trains. Incorrect categorization rates for broadband and frequency banded clicks were 0.02 (SD 0.01), but only 30% of the click trains met the categorization threshold. To increase the proportion of categorized click trains, model predictions were pooled within acoustic encounters and a likelihood ratio threshold was used to categorize encounters. This increased the proportion of the click trains meeting either the broadband or frequency banded categorization threshold to 98%. Predicted species distribution at the 30 study sites matched well to visual sighting records from the region.

Combining whistle acoustic parameters to discriminate Mediterranean odontocetes during passive acoustic monitoring.

Acoustic observation can complement visual observation to more effectively monitor occurrence and distribution of marine mammals. For effective acoustic censuses, calibration methods must be determined by joint visual and acoustic studies. Research is still needed in the field of acoustic species identification, particularly for smaller odontocetes. From 1994 to 2012, whistles of four odontocete species were recorded in different areas of the Mediterranean Sea to determine how reliably these vocalizations can be classified to species. Recordings were attributed to species by simultaneous visual observation. The results of this study highlight that the frequency parameters, which are linked to physical features of animals, show lower variability than modulation parameters, which are likely to be more dependent on complex eco-ethological contexts. For all the studied species, minimum and maximum frequencies were linearly correlated with body size. DFA and Classification Tree Analysis (CART) show that these parameters were the most important for classifying species; however, both statistical methods highlighted the need for combining them with the number of contour minima and contour maxima for correct classification. Generally, DFA and CART results reflected both phylogenetic distance (especially for common and striped dolphins) and the size of the species.

Automated segmentation of linear time-frequency representations of marine-mammal sounds.

Many marine mammals produce highly nonlinear frequency modulations. Determining the time-frequency support of these sounds offers various applications, which include recognition, localization, and density estimation. This study introduces a low parameterized automated spectrogram segmentation method that is based on a theoretical probabilistic framework. In the first step, the background noise in the spectrogram is fitted with a Chi-squared distribution and thresholded using a Neyman-Pearson approach. In the second step, the number of false detections in time-frequency regions is modeled as a binomial distribution, and then through a Neyman-Pearson strategy, the time-frequency bins are gathered into regions of interest. The proposed method is validated on real data of large sequences of whistles from common dolphins, collected in the Bay of Biscay (France). The proposed method is also compared with two alternative approaches: the first is smoothing and thresholding of the spectrogram; the second is thresholding of the spectrogram followed by the use of morphological operators to gather the time-frequency bins and to remove false positives. This method is shown to increase the probability of detection for the same probability of false alarms.