Classification of Risso’s and Pacific white-sided dolphins using spectral properties of echolocation clicks
Map of study area and delphinid recording locations offshore of southern California. Coastline, , and bathymetric contours are represented. The inset expands the cluster of recordings from San Clemente Island area. This cluster represents increased effort, not increased presence. (★) FLIP location represents multiple sightings from Table V. (●) Delphinus delphis, (◼) Delphinus capensis, (◆) Grampus griseus, (▼) Lagenorhynchus obliquidens, and (▲) Tursiops truncatus.
Example wave form and corresponding Teager energy of a Pacific white-sided dolphin click. Note the reverberations present in the wave form. (A) The click wave form, (B) the click wave form with denotations of click end points and data points above Teager energy noise floor threshold, (C) the Teager energy of the wave form, the running mean of the Teager energy, denotations of click end points, and data points above the noise floor threshold. The solid vertical bars represent the time range of the complete click peak.
Example spectra and wave forms for (A) Delphinus capensis, (B) Delphinus delphis, (C) Grampus griseus, (D) Lagenorhynchus obliquidens, and (E) Tursiops truncatus.
Concatenated spectrograms and mean normalized spectral plots of complete clicks for each species using Hann-windowed data. (A) Delphinus delphis, (B) Delphinus capensis, (C) Grampus griseus, (D) Lagenorhynchus obliquidens, and (E) Tursiops truncatus. For the concatenated spectrograms, frequency is plotted on the -axis. Click number, rather than continuous time, is plotted on the -axis. The magnitude of the frequency content is represented by the color such that lighter blue represents greater magnitude. Alternating high and low amplitude spectral bands are apparent between 20 and for . griseus and L. obliquidens. The black vertical lines in spectrograms represent breaks between recording sessions. For the mean normalized spectral plots, the solid line represents the mean and the dotted lines represent one standard deviation.
Histograms of frequency values of spectral peaks (left) and notches (right) for (A) Delphinus delphis, (B) Delphinus capensis, (C) Grampus griseus, (D) Lagenorhynchus obliquidens, and (E) Tursiops truncatus. Each bar represents one FFT frequency bin. The dotted lines represent the upper boundary of estimated histogram background noise as determined from peak and notch randomization procedure. Groups of bars that rise above this line represent consistent peaks or notches that are analyzed further.
Univariate Gaussian mixture model fits to spectral peak and notch histograms from the training data set for Grampus griseus (A) peaks and (B) notches and Lagenorhynchus obliquidens (C) peaks and (D) notches. Each bar represents one FFT frequency bin. Mean and standard deviation estimates of the Gaussian distributions of consistent peaks and notches are noted.
Concatenated spectrograms and mean spectral plots for (A) subset A and (B) subset B of Lagenorhynchus obliquidens clicks using Hann-windowed data. The black vertical lines in spectrograms represent breaks between recording sessions. The consistencies of spectral peaks and notches across recording sessions of each subset are apparent, as are the distinctions between clicks from the two subsets, revealing that Lagenorhynchus obliquidens produce two distinct click types.
LTSAs of data from seafloor HARP instruments show echolocation bouts, which exhibit similar spectral peak/notch structure to that found for (A) Grampus griseus and (B) Lagenorhynchus obliquidens, including both the (type B) and (type A) clicks. The first half hour has a group producing clicks of type B. In the second half hour, a group begins producing type A clicks while the type B clicks fade out. The vertical line represents the beginning of type A clicks. Each plot shows of data.
Published click characteristics of common, Risso’s, Pacific white-sided and bottlenose dolphins.
Survey and instrumentation information. Frequency response and gain of the acoustic instruments are described in detail in the text. Surveys conducted from stationary or drifting platforms are indicated by a blank field under tow distance. Abbreviations: CC: CalCOFI oceanographic survey; SC: southern California instrumentation survey; SCI: San Clemente Island survey; FLIP: FLIP moored observations.
Summary of data included in click analysis. Survey platform, numbers of clicks, click trains, and school sizes are presented for each recording of each species. Each recording session represents a new school of dolphins. Abbreviations: Dc: Delphinus capensis; Dd: Delphinus delphis; Gg: Grampus griseus; Lo: Lagenorhynchus obliquidens; Tt: Tursiops truncatus. CC: CalCOFI oceanographic survey; SC: southern California instrumentation survey; SCI: San Clemente Island survey; FLIP: FLIP moored observations.
Means and standard deviations of local peaks and notches for Grampus griseus (Gg) and Lagenorhynchus obliquidens (Lo). Means from training data were estimated from Gaussian mixture models. Means from testing data represent frequency variability of the value of the peak or notch used for ANOVA analyses. The percentage of total clicks containing a peak in this frequency range is also presented.
Results of nested ANOVAs testing for variation in peaks and notches between species (Pacific white-sided and Risso’s dolphins) and among recordings nested within species. Significant differences are indicated by asterisks. ANOVA tests among multiple comparisons, such as among recording sessions, only indicate that significant differences exist, not which sessions were different. See text on post hoc analysis for further detail of the differences between recording sessions.
Subsets of Pacific white-sided dolphin recording sessions as distinguished by Tukey’s post hoc analyses. While different surveys used different recording gears, subset A includes recordings from all surveys, including FLIP, while subset B only includes recording sessions from FLIP. This suggests that differences between recording sessions are not a function of the differing gear used but rather represent two distinct click types produced consistently within a dolphin school as described in the text. Sessions 13 and 21, also recorded from FLIP, were not significantly different from either subset and visual inspection suggests that they contain both click types.
Article metrics loading...
Full text loading...